the origin of thurneysen's law

THE ORIGIN OF THURNEYSEN'S LAW

A detailed analysis of the evidence

1. Introduction

Herewith are presented the results of the investigation into Thurn-eysen's law in Gothic alluded to in my earlier articie (Woodhouse1998) on early Germanic obstruent development. For a good intro-duction to Thurneysen's law with literature, discussion of a numberof key issues, including criticism of recent views, the reader is re-ferred to Suzuki (1992). Sonne criticisms of Suzuki's own views arecontained in my earlier (1998) paper.

It is clear frorn the ongoing controversial Status of Thurneysen's lawthat no account of the law published hitherto has subjected the dataon which the law is based to a thorough case-by-case analysis in orderto determine exactly which of the items actually require Thurneysen'slaw for their explanation and which can be adequately explained byother means, notably Verner's law. Thurneysen's richly supported sem-inal paper (1898), demonstrating the curious regularity with which asubstantial body (around a hundred items) of Gothic data compliedwith the statable principle of voicing dissimilation that now bears hisname, was intended äs a rebuff to a Statement by Johannes Schmidt(quoted by Thurneysen 1898, p. 208) treating the phenomenon ästhe - apparently direct - outcome of Verner's law. Thurneysen's con-clusion (p. 214) that his Gothic data could not be used to determineaccent position in Proto-Indo-European (PIE) is, at least äs thingsstand, perfectly correct, but this does not of course mean that the, soto speak, bona fides of the Gothic material cannot be usefully testedby confronting it with Information about accent position in PIE thatcan be ascertained from other sources. Attempts to do this have beenmade, but, äs has been indicated, they have for the most part concen-trated on generalities or, äs in the case of Bammesberger (1986b),dealt with only a limited section of the corpus. This is hardly surpris-ing, given the quantity of data involved. Even Flickinger (1981), whoproduced some genuine and useful insights (cf. 3.0, 3.0.1 below),treated accent position only in relation to some of the Suffixes and some

Brought to you by | University of Queensland - UQ LibraryAuthenticated

Download Date | 9/28/15 5:14 AM

188 ROBERT WOODHOUSE

of the stems s separate entities, regretting that whole words cognatewith the Gothic items were hardly attested at all in Vedic Sanskrit.The present study is intended s a first attempt to remedy these andother deficiencies of studies of this kind.

In order to keep the labour involved in such an undertaking withinreasonable bounds while at the same time Consulting a significantbody of data - and observing that Flickinger's excursion into othersources resulted only in the quaint addition of an alleged »ganisa >sal-vation<« (1981, p. 68; the form cited is 1. sg. pres. act. ind., meaningmostly, s at Mt. 9,21, >I shall be saved<) - I have decided to base thestudy on Thurneysen's original collection, with which it should beread in coi\junction, and to treat the material in much the same se-quence, classifying it however into the explicability groups indicatedabove, viz.: material conforming to Thurneysen's law but explicableon the basis of Verner's; material explicable only on the basis of Thurn-eysen's law; material whose explanation is unclear; and material thatcannot be explained by Thurneysen's law because it contradicts it.For Gothic etymology I have relied chiefly on the well known dictio-nary of Feist (1939) and the update of this by Lehmann (1986), ac-knowledging their assistance simply by mentioning these two schol-ars' names without date, and citing the relevant dictionary article onlywhere this is not obvious from the context or not easily found fromcross references within the works themselves. More recent scholarlyliterature has been scanned for Updates.

Although a residue of uncertainty remains, some of which mustinevitably yield to future research, I believe that the present studybrings considerable clarity to the main outlines of the subject andthus makes possible a properly informed evaluation of the nature ofthe process we call Thurneysen's law.

And so: to the material.

2. Data for alternating sibilants -s-l-z-

For the -is-/-iz- group Flickinger (1981, p. 72) offers curious com-parisons with Skt. rdjani >night<, r ma >dear<. The obvious compari-sons for Goth. (dat.) hatiza >hatred, anger< and (dat.) rigiza >dark-ness< are the neuter -es-stems Gk. (Doric) κάδος >sorrow< and Skt.rajas- >(lower atmospheric region of) vapour, fog, gloom<, Gk. έρεβος>darkness of the underworld<, respectively (Feist, Lehmann), though



THE ORIGIN OF THURNEYSEN'S LAW 189

in Gothic these words have gone over to the PIE -o- (Germanic -a-)declension, äs has happened in other branches of IE äs well, e.g.Armenian and Baltic (Brugmann 1906, p. 520-525). In addition, hatizadiffers from its Greek congener in having presumably either zerograde (*kh2d-) or theme II (^kh^ed-) of the root.

Current theory (see Beekes 1995, p. 185f.) suggests that in PIE theparadigm of the neuter sustantives in -es- once had mobile accent,with accented root syllable in some forms and accented -es- suffix inothers. The e-vocalism that predominates in the root syllable and thesuffix was originally proper only to the accented syllable in any partic-ular member of the paradigm, the unaccented syllables having zeroor, particularly in final syllables, o-grade. The spread of e to both sylla-bles in so many forms, with concomitant fixing of the accent on theroot syllable, is due to levelling which must have begun already inPIE (see examples in Brugmann 1906, p. 517-528).l

Beside the neuter substantives there was also an acUectival para-digm apparently with columnar accent on the -es- suffix, though incompounds the accent position was more variable. These acyectivesoften had active meaning and were also capable of substantivization,äs can be seen by the examples collected by Brugmann (1906,p. 528f.), e.g. Ved. taras- >velocity, energy, etc.< : taras- >quick, ener-getic<, raksas- >act of guarding; something to be guarded against =härm, ii\jury< : raksas- >harmful = an evil being or demon<, etc.

These brief remarks supply the minimum background required tomake a Start on classifying the material for the sibilants. Other theo-retical material can be dealt with äs the need arises.

2.1. Inherited material conforming to Thurneysen's law

2.1.1. Material supported elsewhere in Germanic:2.1.1.1. agisa (neut; dat. sg., nom. pl.) >fear, terror< has Gm. cog-

nates with voiceless -s-, OE egeslic, OHG egeslih >fearsome, dread<,OE eg(e)sa (masc.) OS, OHG agiso, egiso (masc.), OHG egisa (fern.,but this is possibly secondary) >terror<, Norw. egse >Aufgeregtheit<,Norw. dial. egsa >durch Schreck zu etwas treiben, reizen< (Falk/1960, s.v. ekse), etc. The gender Variation (certainly neut./masc.) andthe phonology support Interpretation äs a substantivized acüective

1 Brugmann (e. g. 1906, p. 524) believed that the -e-e- vocalism was char-acteristic of these stems in PIE.




meaning >something dread, provoker of fear<. This meaning is attestedin Gothic (for nom. sg. agis) in Rm. 13,3 and is clearly present in thetwo WGm. acUectives (above), and also in OE egesfull >fearsome,dread< refening to Holofernes in the >Judith< (257) and possibly in OEegesa (ibid. 252) if this is in apposition with the provoker of fear, viz.m&gen Ebrea >the might of the Hebrews<. Beside this there ought tobe an original neuter Substantive *agiz- meaning the emotion of fear,etc., itself. Possibly this form is reflected in OE ege, which in theMercian hymns (Whitelock 1974, p. 192, 195) glosses timor >fear,dread< (>Leid/sorrow< in Feist/Lehmann), meaning, apparently, theemotion. The same meaning is widely attested, however, in the -s-forms äs well: the remainder of the Gothic attestations ofagis- appearto have it - certainly the gen. sg. (agisis) in Jo. 7,13, the cognateinstr./dat. in Lk. 2,9 and cognate acc. in Mk. 4,41, and no doubt thenom. pl. in 2 Cor. 7,5; so too OE Meotudes egsa >fear of God< (Seafarer103); and both meanings are well attested for OHG egiso (Grosse 1971s.v.), an r-form, egiro, in a 9th Century MS being regarded (ibid.) äsan error.2

2.1.1.2. sigis- >victory< also has masc. WGm. cognates; the samesemantic duality can apply äs in Eng. triumph (= both the feeling andthe event). The -s-forms are generally less conspicuous elsewhere inGm. Feist lists a couple of personal names containing the -s- (e. g.Segis-mundus-, these have been discarded by Lehmann). NHG pre-sents a healthy contingent of Sieges-forms.3 The possibility that the-es- in these is genitival in origin is made unlikely by the seeminglycomplete absence of would-be genivital -s-forms among the com-pounds of an otherwise comparable word such äs Hof-. Thus theSieges-words almost certainly attest an originally acüectival -es-stem.4

2 Why the substantivized ac^jectival form has been generalized at the ex-pense of the original neuter Substantive can only be a matter for specula-tion. Evidently, the provocation of fear occupied a more central place inthe world of the early Teutons than the experiencing of it, but perhapsthis was due less to an assumed >Inbrunst< than to a fondness for the alu.

3 Cf. Sieges-aufzug, -bahn, -bogen, -feier, -fest, -gewiß, -Herr, -lauf,-taumel, -trunken, -wagen, -willen, -zeichen, -zug.

4 On the basis of these and other WGm. masc. cognates to neut. -es-stems, and some other very limited data, Brugmann (1906, p. 522 f.,533 f.) thought to identify some PIE -is- stems. Consideration of thismaterial is not without significance for the question of deciding themost reliable source for determining the PIE accent position in evaluat-ing Germanic data. Brugmann's solitary Greek example, viz.




2.1.2. Material with weak or no general support in Germanic:2.1.2.1. haiiza (dat.) >hatred< has the appropriate meaning for an

original neuter Substantive with accent on the root syllable, hence -z-is the expected result of Verner's law. (ON hatr Supports this; Eng.hatred has an extra suffix; in general I have reservations about citinggeneral Germanic -r- s support for Gothic -z-.}

2.1.2.2. riqiza (dat.) >darkness< (reflecting the normal -e-e- vocal-ism): s for hatiza.

2.1.2.3. walisa >genuine, beautiful· (possibly with o-grade in firstsyllable, see fh. 4 above): voiceless -s- reflects the expected acUectivalaccent on the suffix *-es-.5 Bammesberger's (1980) proposal to regard

>dust<, clearly accents the first syllable and has the same o-grade s ispossible in walisa and agisa, whereas the handful of Vedic examplesagree with Gm. -s- by accenting the suffix, thus presenting the sameaccentual split s is observed with the *-as-stems (e.g. Gk. κρέας>meat<, Ved. kravis- >raw meat<). The practical significance of Brug-mann's theory, in my view, is that it lends some support to the ideathat PIE *i, in addition to representing the syllabic allophone of *y,may have been preserved in some cases s an old intermediate stagebetween accented *e and zero, since immediate loss of all unaccentedvowels is a priori unlikely and would lead at a stroke to a language inwhich all vowels were accented (cf. Beekes 1995, p. 166). Moreover,instances in which *i alternates with *e, and also cases where *u al-ternates with *o/o, were documented by Speirs (1984, p. 39f., 57f.,171-176 and passim) partly in order to explain zero grade (p. 67-70),though Speirs's explanation was not really helped by his appeal to anidiosyncratic pair of laryngeals. The objection that PIE *i is never lostin zero grade is not cogent since the retention of this *i is due to itsalternation with PIE *i/ and paradigm coherence. >Intermediate zerograde< i and u are preserved in Baltic (and more or less in Slavic) inthe vicinity of liquids (etc.; Stang 1966, p. 33-35), and also in Sanskrit,if they were lengthened in this environment by a nearby laryngeal, andpossibly in some other languages s well (Speirs 1984, p. 163-170,etc.). The theory that the quality of these vowels was determined bythe preceding consonant is now rejected for Sanskrit by Mayrhofer(1986-1996, s.v. GAR1 1) and is equally futile for Baltic-Slavic (seeKortlandt 1978 a, p. 240). If, therefore, Brugmann's -is- stems representa few instances in which a near zero grade i has been preserved, ssometimes also happens, for reasons >still unclear< (Beekes 1995,p. 107), in Hittite in these stems, e. g. nepis >sky<, these would consti-tute another case in which, s in Ved. matar-, etc., OHG muote/ir, OEm do/er : Gk. μήτηρ, acc. μητέρα >mother<, Germanic agrees betterwith Vedic than with Greek on the position of the accent.

5 The suggested (Feist, Lehmann) protoform *waliz should be amendedto *waliz- >Wahl< and *walis- >der Gew hlte<.




the word s an ex-comparative is unconvincing for the followingreasons: (1) the absolute use of comparatives cited in support evi-dently dates from a later epoch than walisa, if the latter is indeed alsoa comparative, since the forms cited, viz. framaldrozei and managizo(Bammesberger 1980, p. 3), have either resisted Thurneysen's law orhave eliminated its effects by analogy, and also walisa would then bethe only comparative form in the language with ending -is-; (2) thepersonal names OE Weisung, OHG Welisung (Feist, Lehmann) ap-pear to contain the exact cognate of walisa, yet Thurneysen's lawcan hardly have applied to them; (3) the objections to the -es- stemetymology of walisa (see Bammesberger 1980, p. 2 f.) are by no meansinsurmountable, since (a) levelled β-vocalism of the root is not obliga-tory, nor do deviations from e-vocalism necessarily have to be of PIEage,6 and (b) the conclusion that walisa was declined only weak, eventhough it is somewhat insecure, being based entirely on four similarvocative expressions,7 nevertheless sits well with the fact that walisa,if it means >real, genuines belongs with a small group of qualifiers

6 Cf. (i) attesting zero grade: Skt. uras- >breast< (Wackernagel/Debrunner1954, p. 227; Mayrhofer 1986-1996, s.v.), Gk. (neut.) κύτος >receptacle,cover; protuberance, cavity< (of PIE age if identical with Lith. dial.k tas >hem of garment<, Illich-Svitych 1979, p. 37); Lith. puvesis >rottenwood<, griuvesiai >ruins< (ibid., p. 34); in Germanic, OE stulor >theft<(Wright/Wright 1925, p. 213); and (ii) attesting o-grade: apparently in-herited in Lal.foedus >league, alliance<, but said to be transferred fromo-stems in Lat. pondus >weight<, modes-tus >moderate; modest< (Leu-mann 1977, p. 378) and also in Lith. l pas >leaf< beside Gk. (neut.)λέπος >pod< (Illich-Svitych 1979, p. 37 [»PIE lopos« is presumably amisprint for lepos])-, in Germanic cf. OFris. kl th, kl th, pl. kl ther>dress<, OE cl d, clxd (cf. Steller 1928, p. 45; Holthausen 1934, s.v.),since there appears to be little or no support for a root form **glh&eyt-(cf. also Fraenkel 1962-1965, s.v. glieti)\ OHG (neut.)/a?ta)/i, pl.farhir >pig<, (possibly) Lith. pafsas >piglet<, beside Lat. (masc.) porcus>pig<, especially if, s seems likely despite the misgivings of Walde/Hofmann (1938-1954, s.v. 2. porcus), the word belongs with *perk->speckled<; probably also OE sweng >blow<, perhaps spere >spear< andsome other items with OE stem vowel ~e- recorded by Wright/Wright(1925, p. 213). Cf. also Hamp (1973, p. 142 f.).

7 See Bammesberger (1980, p. 1) for these and also for the one otheroccurrence of the word, viz. in the phrase weihans j h walisans; inthis phrase the weak form walisans is necessarily in tandem with thatof weihans (an ac^jective also possessing a strong declension) and isthus, strictly speaking, non-probative with respect to the ability or in-ability of the parent a^jective to generate strong forms.




possessing only weak declensions and indicating not so much descrip-tive attributes äs a relationship of the target item with respect to other(potentially) similar items, viz. ibna >equal<, ainaha >sole, solitary,only<, sama >the same, i. e. not some similar item<, taihswa >the right<(äs opposed to >left<, Goth. tdeiduma, significantly, a comparative),hence, at some point in its semantic development, walisa * >the genu-ine one<.8

2.1.2.4. l)ewisa (voc. pl.) >servants<. If this is an -es- derivative of*iekw- >run< (Lehmann) the meaning is incompatible with that of anoriginal neuter Substantive, pointing instead to a substantivized neuteracUective (on the gender cf. the Old RUSS. neut. otroca >servant<), thelong root vowel possibly being that of a vrddhi collective, äs Brug-mann (1906, p. 527) suggested. Alternatively, the word may have theaccented *-wes- of the perf. act. participle (Feist citing Walde; cf.Streitberg 1963, p. 213 f.) with the same root vocalism äs in berusjos(cf. 2.1.2.8 below). Either way this word, with its medial -w-, belongswith the distinguished group of Gothic words preserving the voicingeffects of Verner's law.9 This raises certain questions concerning thechronological connection between Thurneysen's law, the arrest ofVerner's law voicing and some other issues. These will be dealt within the concluding section 6.2 below.

2.1.2.5. riqizjan >grow dark< is a denominative. The -z- agrees withthe accent on the suffix -ya- of denominatives in Vedic, where, coinci-dentally, the commonest consonant stems giving rise to denomina-tives are those in -äs- (PIE *-es-) (Macdonell 1968, p. 401).

2.1.2.6. hatizofj (2. pl. pres., not 3. sg., pace Feist, Lehmann) >areangry<, a denominative formation of mixed derivation (Prokosch 1939,p. 157-159; Wright 1954, p. 156; Streitberg 1963, p. 312) making diffi-cult or impossible any rigorous Statement on accentuation. True, Bam-

8 The above list is given by Streitberg (1920, p. 129 f.) along with someother material that requires a different explanation: alewjo >of Olives<,a local place name, so always definite, and unlveilo >intense<, a hapax,like UuLshandja >empty-handed<.

9 On two grounds, therefore, the protoform given by Lehmann may besupplied with an accent, thus: *tekwes-o~. Note: (1) the effect of vrddhion the accent of collectives appears to be unpredictable, cf. sapta- andsdpta- >group of seven< (Brugmann 1906, p. 648), both occurring twiceeach in the Rigveda (1,20,7; 2,19,7 and 8,55,5; 8,59,5, respectively), de-spite Mayrhofer (1986-1996, s.v. sapta)', (2) Brugmann's (1906, p. 527)appeal to an unaccented suffix *-wes- can be discarded.




mesberger (1986 a, p. 38 f.) would derive Gothic verbs of this classfrom PGm. forms continuing accented *-ye-/-yo- suffix, which wouldindeed provide a natural Verner's law explanation for the -z- in ourform, except that there appears to be no evidence that the Gothicforms ever contained this suffix (cf. also ibid., p. 84, especially on thedifficulty of trying to reconstruct a *?/-suffix in the verbs in -β-)· Onthe contrary, a few accented denominatives lacking the characteristicaccented -ya- suffix occur also in Vedic, viz. bhisakti, krpananta,and possibly vananvati (Macdonell 1968, p. 401), and a reasonableworking hypothesis based on them is that the accent of the verbtended to be that of the noun. On this basis hatizofr would naturallyhave the consonantism ofhatiza (cf. 2.1.2.1. above).

2.1.2.7. swartiz(l)a (dat.) >ink<. The -z- is due to Verner's law, cf.Zimmer's (cited Feist, Lehmann) comparison with the suffixes of Ved.tam-isra-m and tam-asa-m, both meaning >darkness<.

The next two items offer further valuable insights into early Ger-manic treatment of perfect active participles.

2.1.2.8. berusjos (masc. pl.) >parents<, representing in origin the fern,of the perf. act. participle with levelled accented suffix *-us ( s inboth Vedic and Greek) + (fern.) -ϊΐ-iya, or laryngeal equivalent (possi-bly, with Feist, originally ~iy , but see next item 2.1.2.9), and adaptedto masc. gender either by semantic default or by reanalysis of fern,(possibly neut.) dual s masc. pl. (cf. Bammesberger 1995, p. 6), thearticle fxii being capable of supporting all three possibilities (= Skt.te which represents nom./acc. fem./neut. du. and masc. pl., but notneut. pl., pace Lehmann). The vrddhi of the root occurs also in thepret. pl. berum (= PIE perf.) (Feist, Lehmann; cf. also frewisa, 2.1.2.4.above).

2.1.2.9. *jukuzi (fern, i-declension) >yoke<: suggested origin fromfern. perf. act. participle rejected by Feist, but considered possibleby Lehmann based chiefly, no doubt, on the able demonstration byBammesberger (1965; cf. also Bammesberger 1986 a, p. 103). In orderto account for the apparently aberrant accentuation of the Gothicword we begin by observing that the ablaut of the fern. perf. act. parti-ciple Ved. vidusi, Gk. Ιόνια shows that the accent in these forms isnot original, but must once have been on the fern, termination(s),which in turn must also represent some levelling consequent perhapsupon the settling of the accent in its attested position. With jukuziBrugmann (1906, p. 544) compares OHG feminines chilburra >ewe-lamb<, zaturra >meretrix<, chuburra >boat, raft<. Phonologically speak-




ing, all four of these Germanic words can be derived unproblemati-cally from the levelling of protoforms in -us-e^h^l-us-yeh^. At least inthe case of the first three (i. e. excluding chuburra), this derivationclearly meets with no semantic problems, since the perfect indicatesa settled characteristic, äs in the Vedic/Greek example(s) above,which cover such meanings äs >knowing, wise, sensible<. The fourthitem, chuburra, probably belongs with the group listed under the rootgeu-, etc., by Pokorny (1948-1959, p. 393-398), cf. e.g. OHG kiol, etc.,>ship<, kiuUa >pocket< (p. 397), Swed. kypa >round receptacle made ofstraw<, MLG kumme >deep, round, vessel< (p. 396), OE cofa >chamber,hiding place, hollow<, OHG chubisi >hüt, cottage< (p. 395) (see alsoBeekes 1996, p. 223-227).10 The basic meaning thus seems to be>structure that maintains a hollow, rounded shape<, which is an idealmeaning for a perf. act. participle. These four Germanic words canthus represent archaisms which early became separated from the par-ticipial paradigm and thus remained immune from the levellings ofaccent and other innovations that were clearly under way if not com-pleted in the living paradigm during the PIE period (pace Bammesb-erger 1986a, p. 103 and 161, n. 9, who equally clearly believes thelevellings occurred independently in some daughter languages but notin Germanic).

2.1.2.10. riqizeins >dark< had accent on the suffix following the -z-,whether this be *-mo-, so accented in both Vedic and Greek, or theless certainly attested -eino- (Brugmann 1906, p. 275-277).

10 This last may show the sort of realignment to an -es- suffix that somehave posited (unnecessarily) for OE gycer >yoke< (e. g. Walde and Po-korny: see Bammesberger 1965, p. 417 f.) except that OHG chubisi at-tests in addition the characteristic accent place of an (eventually sub-stantivized) acUective of this formation. Such a realignment could havebeen due either to Speakers' growing unfamiliarity with the derivationor, with equal probability, to a fading need to keep the precise deriva-tion of a familiär item alive. Cf. the confusion that eventually descendson hackneyed phrases which results in slips like sense offalse securityforfalse sense of security and probably also in the common confusionof vaguely synonymous and phonetically overlapping items such äs:Eng. suppress, repress and oppress; the common Australien malaprop-ism mitigate (sc. militate) against\ and the like.




2.2. Items whose consonantism appears to be determinedby Thurneysen's law

2.2.1. aqizi >ax<. Germanic >cognates< (Feist, Lehmann) suggest athorough confusion of Lat. ascia >ax, mattock, hoe, trowel< with acus>needle< and even ac tus >sharp, cutting<. The context in Luke is anagricultural one, so that the old Germanic word for battleax (OHGbihaly Eng. bitt) was not appropriate: Speakers of early Germanic evi-dently favoured keeping the two concepts apart. The -s- elsewherein Germanic suggests that the Gothic word is a genuine example ofThurneysen's law.

2.2.2. barusnjan >regulate with religion, etc.<: despite the uncertain-ties of meaning, structure and etymology, the -ja- suffix indicates ac-cent after the -s-, which must therefore owe its existence to Thurn-eysen's law.

2.3. Items whose causal relationship with Thurneysen's lawcannot be determined

2.3.1. rimisa (dat.) >quietness<, hapax, only in the phrase ?m/> rim-isa >quietly<, which, in the context of 2 Thess. 3,12, must contrastwith περιεργαζομένους >busybodying< in the previous verse and thussignify something like >with quiet demeanour< (hardly >[working] withrest<, s Lehmann's gloss might suggest). Lith. r mas (Lehmann) ap-pears to confirm the existence of a root-accented neuter Substantivein -es-, given that fllich-Svitych (1979, p. 35-39) cites four words ofthis structure and origin (par$as, ja kas, kutas, tapas) among his tenexamples of PIE barytone neuters in thuanian but has no -es- stemsamong the eight items he cites for his PIE neuter mobile oxytones; inaddition, the - s desinence augurs well for an old neut. subst. nom. sg.in *-os, rather than the *~es of the a^jectives. However, the LithuanianSubstantive is not well known to the world of learning. It figures notat all in the five volume dictionary by Niedermann et al. (1932-1968).According to A. Kurschat (1968-1973, s. v. raraas), who like his fatherbefore him (F. Kurschat 1883, s.v.), records the word without accent,it was heard in a folksong in the middle of the last Century by Nessel-mann, who went so far s to indicate stress on the first syllable of thecitation form. Lehmann's circumflex was apparently supplied, perhapsunder the Stimulus of the word's possessing an IE etymology, byFraenkel (1960-1962, s.v. r mas 2), who also cites Nesselmann s his



THE ORIGIN OF THURNEYSEN'S I.AW 197

source but neglects to indicate precisely the all-important accentualparadigm. Thus independent confirmation that our Lithuanian wordbelongs to the old barytone accent class (2) is lacking.11 Against this,Gk. (compar.) ήρεμέστερος (Feist, Lehmann) presupposes the exis-tence of an acUective (masc., fern.) *ήρεμής, (neut.) *ήρεμές, and giventhe apparent confusion of subst. and ac\j. in the >terror< area in Ger-manic (cf. 2.1.1. above), a similar confusion cannot be ruled out forthe >calm, quiet< words (note that each of these two English glossesembraces both functions). But although it is tempting to argue thatthe active meaning expected for the acUective, viz. *>promoting peace,conducive to peace<, would lead to a Substantive with a more suitablemeaning for the context, such s *>peaceful ways, a peaceful disposi-tion, a quiet demeanour< and the like, there is no proving that this isany more suitable than plain >tranquillity<. Nor does an examinationof alternative Gothic translations of Gk. ησυχία carry us any furtherforward.12 Consequently, both the available alternative interpretationshave their attractions and their problems, so that a decision favouringone over the other appears impossible.

11 The same is, however, tantalizingly not true for the homographic loanmeaning >piledriver< or >pillar< recorded a^jacently, also without con-sensus over the accent, by both Fraenkel and A. Kurschat, and suppliedby the latter with a plural form indicating class (2). While the possi-bility of mutual influence between the two words cannot be entirelyruled out, there is no clear evidence for it in our sources.

12 Beside Goth. rnify rimisa translating Gk. μετά ησυχίας are Goth. inhaunifxii (A; B has in hauifxii) >in/with humility(?)< and in fxihainai>in silence<, both translating εν ήσνχίφ (l Tim. 2,11 and 12) indicatingSt. Paul's desperately pessimistic view of how a woman should disportherseif in any intellectual dealings with a man. The stylistic Variationevident here in the Gothic reappears in the Latin text of l Tim. within silentio vs. in taciturnitate (cf. Friedrichsen 1939, p. 128-135, whoneglects, however, to mention the case of ησυχία), whereas the differ-ence in the preposition in the Thess. context (Gk. μετά, Goth. mi ,Lat. cum [silentio}) rests clearly enough on the difference of meaningsignalled in the handbooks for the Greek text (e. g. Bauer 1958, s. w.μετά III, col. 1007, et εν ΙΠ.2, col. 517; Schirlitz 1908, sw. μετά I.d,p. 264, et εν 3.a, p. 140; etc.), viz. the phrase with μετά indicates acircumstance accompanying an action, i. e. it refers to general ambientconditions applicable to all participants, while the phrase with εν indi-cates the state or manner in which a particular person acts and so isoften reducible to a simple adverb of manner, e. g. εν τάχει >quickly<,and can thus easily refer to the unilateral behaviour of a single partici-pant in the action, in this case the woman.




2.3.2. walwisoda (pret.) >(intrans.) roU about, roll to and fro<: an-other denominative of the type of hatizofr (cf. 2.1.2.6 above). It istempting to speculate that while the semantics naturally favour thederivation of *hatizon >be angry< from the neuter abstractum hatiz->anger<, *walwison would equally naturally derive from a substantiv-ized a^j. *walwis- >rolling, roller-like; roller, sth. that rolls/is rolled toand fro< (possibly some unknown item of the Gothic toolkit), preserv-ing the latter's consonantism - for this property cf. aw iudon >thank<from awiliuda (pl.) >thanks< despite Hufan >sing praise> (and despiteThurneysen's law, unless this is a compound). On the other hand onecannot help noticing that four other verbs of this class with root finalspirants fit the voicing dissimilation pattern, even though they havenothing to do with Thurneysen's law, viz. lafon >invite<, hxirbon >goabout<, sidon >practise< and swiglon >pipe<.

2.3.3. halisaiw (Lk. 9,39) >hardly (ever), almost never<: no judge-ment possible in view of uncertain etymology (Feist, Lehmann) anduncertain Status, the word (word group?) being a hapax in a singleMS not free from scribal errors in the immediate vicinity - cf. (sunu)for sunau (v. 38), (hidrei) for hidre (v. 41), etc. (Streitberg 1950,p. 131-133); and even uncertain componential analysis: if sunsaiw>immediately< contains comparative suns >soonest<, halis- in principlecould also be a fossilized comparative with -s- then due to Thurn-eysen's law. The meaning would then be >least< - and the semanticconnection between >weak< (MHG hei, etc., Holthausen, cited Feist,Lehmann), >thin< (cf. Arabic dcfif- [Classical] >weak, frail< > [vernacu-lar] >skinny, slim<) and >small< (NHG mager, Eng. meagre) is not hardto establish.

The next three words contain the suffix combination -s/z-na-, theaccentual properties of which appear to be variable: root accent inVed. matasna- >paired internal organ in the ehest (lung?)< (cf. matsya->fish<), barytone in karasna- >forearm< (cf. kam- >doer, hand<), oxy-tone in vadhasna- >weapon, Indra's thunderbolt< (cf. vadha- >slayer<),all masc. or neut. Since the Gothic words are all fern, we should per-haps compare Ved. jyotsn - >moonlit night<, Skt. mrtsn - >loam, finedust< (probably oxytone in view of ablaut and [late] Ved. m rtsna->pulverized<), barytone in Gk. (Att.) σελήνη, (Lesb.) σελάννα >moon<.The sparse indications of derivatives mimicking the accent of the sim-plex seem to be offset by a distinct tendency towards oxytonesis,judging by the forms cited by Wackernagel/Debrunner (1954, p. 927f.,though mrtsna- on p. 928 appears to be an error), Brugmann (1906,




p. 282) and Macdonell (1968, p. 133f.). This renders insecure the fairto mediocre arguments that can be mounted to reconcile the conso-nantism of the first two of the Gothic items below with the likely PIEaccent of their respective simplicia and to support the Operation ofThurneysen's law in the case of the third.

2.3.4. artvaznos (fern. acc. pl.) >arrows<: Germanic cognates -OIcel. pr, OE earh - point to root accent of the base, but the Gothicarrest of Verner's law voicing makes it impossible to be certain thatthe -tv- of the compound and, by implication, the -z- of the suffix areoriginal. On the chronological implications this raises, see 6.2 below.

2.3.5. filusna >great extent, multitude<: despite the corruption of thetext of Neh. 5,18 (cf. LXX 2 Esdras 15,18), meaiüng and structureseem to be clear enough. An argument for inherited Status is äs fol-lows. The Suggestion that PIE *i (= Goth. i) might represent reducedgrade of PIE *e (cf. fh. 4 above) raises a doubt that the simplexfilucontains füll grade of the root, given the curious underrepresentationof this elsewhere in IE outside Germanic. The few examples that canbe cited mostly refer unmistakably to liquids and pouring (see Po-korny 1948-1959, p. 798-801) and can largely be eliminated fromconsideration by adopting Fraenkel's (1962-1965, s.v. pilve) Separa-tion of the >marsh< words from the >pour< group. Old Irish ü can repre-sent *pilu- just äs easily äs *pelu- (Thurneysen 1946, p. 38, 47 f.). Ved.parinas- >abundance<, parlman- >id.<, which are claimed äs füll gradederivatives of the >fill/full< root by Mayrhofer (1986-1996, s.w.), haveno correspondences outside Indo-Iranian, so that nothing preventstheir Interpretation äs pari (prefix = adv.) >all around< (= Gk.>around; exceeding[ly]<) + i >go< -»- suffixes, the structure being verysinülar to that of Old Church Slav. obilije >abundance<, viz. ob(i)>around< -»- i >go< + Suffixes.13 Goth. filusna may therefore reflect thegrade and accent of Gk. , Ved. puru-14 and thus have inheritedconsonantism. But since this possibility entails a revision of currentablaut theory it must be either discarded or reworked or set aside forthe future.

2.3.6. hlaiwasnos (fern. nom. pl.) >graves<: differs in suffix vocalismfrom the root-accented -es- stem attested for WGm. by OE ld%w, OHG

13 Of no particular attraction is the alternative based on *vey/w >desire,delight, pursue< offered by Vasmer/Trubacev (1986-1987, s.v. obiVnyf).

14 Beekes's (1995, p. 198) perplexity regarding the PIE paradigm reflectedin these words may thus point to persistent inadequacies in our theo-retical appraisal of PIE vocalism.




(h)leo, (solitary nom. pl.) lewir >grave-mound< (Nedoma 1993 a,p. 124). The Lithuar an cognate sleivas >lame< offered for the simplex(Feist, Lehmann) can teil us nothing about PIE accent since Lithua-nian acjj ectives of this type underwent generalization of accentual mo-bility during the brief historical period ( lich-Svitych 1979, p. 43).

2.4. Exceptions to Thurneysen's law

2.4.1. barizeins >made of barley<: s for riqizeins (cf. 2.1.2.10above).

2.4.2. ubizwa >porch/vestibule of a hall<: PIE *up-es-w (Lehmann,citing Hirt) seems eminently suitable, with *w-' s in Skt. takva- >swiftcreature(?)<, Boeot. κaλfός >beautiful<.

3. Data for alternating dentals -f>-/-d-

3.0. First it is necessary to eliminate some material that clearly doesnot belong here. Thurneysen (1898, p. 211) himself correctly excisedthe weak preterite suffix(es) from the material for his law, butclaimed that levelling had been at work in the case of the Suffixes-i/>a and -Jbro.15 The latter group, which actually includes sundro >es-pecially<, has long since been regarded s having its consonants deter-mined purely and simply by Verner's law (Feist, Lehmann a/ja/rro;also, somewhat incompletely, Flickinger 1981, p. 72).

3.0.1. The case of -i)xL/-ida is particularly instructive. Thurneysen's(1898, p. 211) description of the distribution of the forms is unim-peachable: -ida is found in the only two items in the corpus in whichthe consonant just before the suffix is φ/ and in the solitary hapaxvariant weitwodida, and nowhere eise. Thurneysen plausibly suggestsdissimilation of the double fr sequence for the first two (accepted byFeist wairfnda, Lehmann αιφ$), and confusion with the participlefor the hapax.16 After listing 25 -ifxi items with presuffixal voicedconsonants conforming of course to his law, Thurneysen then cites

15 Indeed this was Thurneysen's (1898, p. 210) explanation for all deriva-tional and flexional Suffixes that contradicted his law, a position re-jected nowadays (Suzuki 1992, p. 32 f.).

16 In the context f>o weitwodida is (Jo. 3,32 in Sk. 4,20 f.) it would thenbe neuter plural and more or less make sense: >the things testified of(= by) him<.




no less than six items with ~ipa and presuffixal voiceless consonants(armahairtilxiy etc.) that of course do not. This shows unmistakablythat Thurneysen's law äs it is usually formulated does not apply tothis suffix.

Such is the force of tradition, however,17 that Suzuki nearly a Cen-tury later (1992, p. 30) was able to content himself with the Statementthat »the suffix -i/w shows a greater degree of exceptionality (than-aha, R.W.)«, which he endeavoured to explain on the basis of thissuffix's greater productivity (1992, p. 43). The productivity argumentthus appears to be without foundation.18

The immunity of the -ifxi- suffix from Thurneysen's law in Gothicis hardly surprising, given the connection established by Flickinger(1981, p. 71) between this fact and the fact that in the bulk of theVedic material containing a related suffix (or Suffixes) the accent fallson the syllable just before the dental consonant of the suffix.19 Per-haps Thurneysen was right to talk about levelling after all, since ifpost-Verner pre-Thurneysen Gothic contained any naturally derived-ida forms, their number would have been so small that they wouldreadily have given way to the overwhelming abundance of -ifxi forms,

17 Cf., e.g., Wright (1954 [= 1910], p. 175), with no new evidence in sup-port of his claim that »-ifxi generally became -ida by dissimilationwhen the preceding syllable began with a voiceless consonant«.

18 Even if Gothic -/ri- is due to levelling, it nevertheless Stands outsidethe material for Thurneysen's law since it is attached without interven-ing vowel to the final consonant of the root. In this position it is verylikely to have simply undergone the usual loss of Verner voicing thatwas typical for that position, especially äs Rigvedic and other evidencesuggests that the accent feil on the *-ti- at the earliest period (Wacker-nagel/Debrunner 1954, p. 631). Thus, in common with the purelyVerner-derived comparative suffix -iz- (Flickinger 1981, p. 72), this-pi- suffix offers no support, or at least no relevant support, to theproductivity-levelling theory either (pace Suzuki 1992, p. 32 f. and 43,fh. 14).

19 Collections of data in both languages - Gothic and Vedic - reveal a10% exception rate. Thurneysen cites 31 Gothic items with -t/ againstthree with -ida. Macdonell (1968, p. 138) and Brugmann (1906, p. 416-418, 451) cite about twenty Vedic items with accent on the vowel imme-diately before the related Vedic suffixes -tä, -tät(i) (even when thisvowel is zero grade, e.g. sü-nr-tä- >gladness<) against two (cited byBrugmann) which atypically preserve the accent position of their baseforms, viz. a-wra-tä >lack of sons<, astatdti- >home<.




leaving behind the two unarguably dissimilated forms attested in ourrecords (and perhaps a very few others) s their memorial.20

It can of course be argued that the dissimilation evident in the twogenuine -ida items testifies to a feeling for well-formedness similar tothe one that I shall be suggesting was responsible for Thurneysen'slaw proper and is thus supportive of the basic thesis being developedhere. The potential this material thus has for shedding its own pecu-liar light on the problem no doubt warrants its continued inclusion indiscussions of Thurneysen's law, but it is clear that the distribution ofthe forms of this suffix in Gothic can be of no further statistical inter-est here.

One further item belongs with this group, although it has tradition-ally been separated from it thanks to misinterpretation of an apparentvariant in a corrupt passage, which has caused much puzzlement. Thegenuine attestation of this item is fullifre (gen. pl.) >full moon(?)< atCol. 2,16. For the alleged acc. sg. */wtfi/> the MS (Mk. 4,28) actuallyhas/w#ei/>, i.e. 3. sg. pres. offulljan (trans.) >fill sth. (with), makesth. f ll (of)<, which is constructed with an >instrumental< genitive.The (corrupt) Gothic text at this point reads:

silbo auk airffa akran bairif): frumist gras, ]xi]yroh ahs, txLJyrohfu eit) kaurnis in /xiraraa ahsa,

which, aping the element order of the (partly forgotten?) Greek, thusrepresents:

>of itself then the earth bears fruit: first the plant, then the ear, thenit [= the earth] fills [it = the plant] with grain in the ear<.ldl

The confusion evident here is mirrored in the Greek text. Apart fromthe plain acc. sg. πλήρη σΐτον of Streitberg's edition, which wouldnormally be taken to mean >(then) [the earth bears] f ll grain<, other

20 The paucity of the data and the nature of the Vedic material entitle usto a doubt which cannot be resolved by other Germanic data, sinceamong the West Germanic cognates of these two items the only formshaving the suffix (see Feist, Lehmann aufo, Lehmann wairf)) all occurin Alemannic OHG: wirthida (Sth Century) >dignitas<, unwirdida/un-wirdeda (lOth/llth Century) >probrum/contemptum< (Graff 1963, vol. I,p. 1019), all pointing to an older levelled wirfrilxi - and it is anybody'sguess which of the two obstruents is to be regarded s the levelledone.

21 For omission of the object pronoun, cf. Mt. 27,48: j h nam swammfulljands aketis >and took a sponge, filling [it] with vinegar<.




versions of the Greek text present the following additional Variante(my source is Ή κοινή διαθήκη, British and Foreign Bible Society,2nd. ed., 1958, p. 112):

(1) acc. with art: πλήρη τον σΐτον >(then f ll [is] the grain [borneby the earth]<, hardly >(then) [the plant] f ll of the grain< with theso called acc. of respect; possibly πλήρη has been misinterpreteds fern. sg. nom. to agree with ή γη >the earth<;

(2) plain nom.: πλήρης σίτος >(then) [there is] f ll grain<;(3) nom. with art.: πλήρης δ σίτος >(then) f ll [is] the grain<(4) nom. plus acc.: πλήρης σΐτον >(then) [the earth, or the plant, is]f ll of grain< - the acc. of respect again.

With such a variety of interpretations present in the Greek text, someuncertainty in the Gothic text at this point is perhaps not particularlysurprising.

Thus only the gen. pl. fulUfe is trustworthy and, in view of OE (incompounds) (-)fyUep(-), ORGfullida >fullness< (Feist), this word canbe safely assigned to the -ijxi group.

3.0.2. After some hesitation, I have decided to join Feist and othersin excluding the three fern, nouns in -frwa - fri(f)apwa >love<, fi-(j)a{nua >enmity<, salifrwos (pl.) >inn, dwelling< (Thurneysen 1898,p. 211) - from consideration, chiefly because of the two neuters in-dw - friwadw >servante< and gaidw >lack< - the first of which (butunderstandably not the second) was cited by Thurneysen (ibid.) s anexception to his law. Since all five nouns in these two groups havevoiced consonants before the suffix there seems to be no basis forclaiming that the difference in their suffix consonants has any connec-tion with Thurneysen's law. Nevertheless a small doubt may remain.The two neuters may be easily disposed of with a reference toVerner's law (Feist, Lehmann, etc.), but the feminines are more prob-lematic. Feist's argument (fijaprwa) invoking Verner's law for the femi-nines on the basis of Slavic accent must now be regarded s out ofdate (see below).22 Moreover, Germanic support for the consonantismof the feminines, which do after all conform to Thurneysen's law, ispatchy. For fri(f)apwa there is nothing at all. For the other twowords, complete loss of -w- in cognates must be assumed: fi(j)apwahas only OE ogafr >hatred<, however there is no dissenting material;salijywos has rather better support, especially in the pluralized direc-

22 Lehmann's uncoordinated emendation excising this has unfortunatelyleft his own position here somewhat unclear.




tional expressions OHG zi selidon, OS te selidun >(towards) home<(Feist, Lehmann). Consequently, some discussion of the backgroundof the suffix in these words is called for.

Somewhat indirect extra-Germanic indications for a generalized ac-cent position for this suffix can be extracted from Indo-Iranian andBaltic-Slavic. The *-twä- suffix is not directly demonstrable in Vedicor Greek. Vedic has root-accented gerundives capable of becoming(neuter) substantives,23 such äs kartva- >to be done; (neut.) task<,hantva- >to be slain< (= OCS [fern.] zgtva >harvest< [*>produce to begathered by cutting/striking<?], etc., Brugmann 1906, p. 448f.). Thefern, of these belong to the -declension, cf. RV 1,161,3: dhenuhkartva >the milch cow (is) to be fashioned<.

Fern. *-twä- substantives occur in Avestan, e.g. Gathic dystvä->teaching, doctrine, dogma<, i.e. *>material or System to be taught/learnt<. The invariable füll grade of the root in these nouns (Wackerna-gel/Debrunner 1954, p. 713) suggests accent on the root syllable, äs inthe Vedic gerundives.

The same accent commonly occurs, äs has been seen, with the suf-fix found, e. g., in Ved. (fern.) priyatä >love<24 so that it seems possibleto regard the -twä- suffix äs representing a blend of the -tä suffixwith the gerundive, complete with the shared accentual propertiesthey exhibit in Vedic. The three formal possibilities appear to meet inLith. \4stvas >chicken coop<, lQSt(v)a >chicken coop, dog kennel, cage<(Fraenkel 1962-1965, p. 377). The Information regarding PIE accentthat can be extracted from the Lithuanian words depends on the ac-centual paradigm class to which they are/can be assigned. In StandardLithuanian, lysta belongs to accentual paradigm class (4), äs doesbrasta >ford< (Kruopas et al. 1972, s.w.), a preference reflected in thetreatment of these two words and their variants in the larger work byKurschat (1968-1973, s.w.), who cites in addition, however, class (2)variants of brasta, brastas, lanstas, Iqstas, Iqstvas and Iqstva (andpossibly some others) from older works. The similarly compendiouswork by Niedermann et al. (1932-1968, s.w., the relevant sections allbeing completed, however, before the death of Franz Bender in 1938,see preface to vol. 2) treats tysta and its variants similarly, but speci-

23 The fact that the metre (and Sievers/Edgerton's law) frequently re-quires the written -tva- suffix to be read äs a disyllable (-tu[v]a-) is ofno consequence in the present context.

24 This disagrees in accent with OE/rfod, freod (fern.) >id.<.




fies only class (2) for brasta and its variant brastva. Thus, relyingadditionally, and with Kortlandt (1978 b, p. 274), on Ülich-Svitych (cf.1979, p. 15, a work originally published in Russian in 1963) for thenormal direction of the transfer being from class (2) to class (4), wecan safely posit the Originals of the all-important variants Igstva andbrastva äs PIE barytones,25 which agrees with the tentative conclu-sions reached above on the basis of Indo-Iranian.26

Some modest support for the Gothic consonantism seems to beforthcoming also in Slavic, though I am aware that the tentative con-clusions that I have been able to draw in this complex field may wellrequire modification. Slavic items with acute (laryngealized) root syl-lables do not count since they can have the accent on this syllableby Hirt's law;27 similarly items like RUSS, lovitva (obsol.) >catching,hnnting<, molitva >prayer< appear to have their stress advanced fromthe (word initial) root syllable by Dybo's law. The same law appears

25 The direction of the transfer agrees with the proposition that the ä-stem barytones comprised a secondary class in PIE anyway, sincestems in ä < *ehz were originally basically oxytones (Beekes 1995,p. 182, 185).

26 Brugmann (1906, p. 450), followed by Wackernagel/Debrunner (1954,p. 714), compared the Gothic (deverbal) feminines with Lith. senätvo>old age< and also with Vedic end-stressed neuters. The comparison ofthese last two types with each other is probably better off without theGothic forms, both phonologically and semanticaüy, since the medialaccent of senätvo seems more certainly to represent a retraction fromthe final syllable (cf. Kortlandt 1977, p. 324) and there is some evidencethat the end-stressed neuters originally denoted a more passive or inertstate than the root-accented feminines, even though the two types con-verged and became interchangeable in Classical Sanskrit. Thus Ved.devatä- means >service to a god, sacrifice<, >divine power< äs well äs>divinity< (Wackernagel/Debrunner 1954, p. 617), whereas devatva-mhas only the latter meaning; Gathic Av. dg^tvä- means >teaching<(cf. 3.0.2 above); further Ved. sünrtd- >generosity<, naristä- >joking,chattering<, iryatä- >alertness<, wratd- >manliness, heroism< (Wacker-nagel/Debrunner 1954, p. 617); vs.janitva-m >state of being a woman<,garbhatva-m >pregnancy<, etc. (Wackernagel/Debrunner 1954, p. 713).And it cannot be denied that >love<, >enmity< and >settlement< can de-note more active concepts than mere senätve >old age<, vienätve >soli-tude, loneliness<, and the like; for >love in action<, e. g., cf. frijafrwos(B: in) frizaiei frijoda uns (Eph. 2,4) >on account of the love withwhich he loved us<.

27 These include (pace Feist fijapwa) RUSS, bitva >battle<, britva >razor<,zatva >reaping<, zertva >sacrifice, victim<, pastva >flock; congregation<(*>to be grazed, fed) (cf. Kortlandt 1975, p. 58-62).




to be responsible for the final stress in RUSS, zratva >guzzling<, whichwould thus be an example of an old accent just before the suffix, ifthe word is old and if its stress is not merely a matter of affectivity.This stress/accent place seems to be confirmed, however, by RUSS.kljatva >oath<, with circumflex root syllable (cf. SCr. kletva >oath,curse<) which would protect it frorn Dybo's law and also, äs far äs Ican see, preclude retraction of the ictus by any other principle duringthe post-PIE period (cf. Kortlandt 1994).

Thus the combined evidence of Indo-Irarüan, Baltic and Slavic sup-ports Verner's law äs the source of voiceless /> in the Gothic suffix-pwa.


3.1.1. Material supported elsewhere in Germanic:3.1.1.1. *framat>eis >alien, allienated, someone else's<, said to have

the same suffix äs in dcUapa (cf. 3.3.4 below). Gm. cognates differin the suffix consonant, OHG fremadi, OS fremühi, OFris. fremetheagreeing with the Gothic word, OE frem(e)de pointing to PGm. */d/(Feist, Lehmann).

3.1.1.2. *framapidans (participle frorn denominative) >estranged<:contrary to expectation, OHG fremidan, OE ä-fremlxin agree withthe Gothic verb in the voiceless form of the suffix (Feist, Lehmann),suggesting either levelling or a late (posterior to the initial onset ofVerner, cf. fh. 40 below) formation within PGm.

3.1.1.3. haimofrli >homestead<, the suffix -ofrli goes back to PGm.*-ol*üa- (Feist, Lehmann).

3.1.1.4. magafrii (dat.) >maiden<: Gm. cognates mostiy confirmGothic consonantism in which -g- is due to Verner's law: OE mag(e)\),OS magath, OFris. megith, OHG magad] only the extended formsOHG magatm (Lehmann), OE maegden, Eng. maiden, etc., dissenting(see also 6.2 below).

3.1.1.5. menofrum (dat. pl.) >month<, OE mönajb, OHG mänöd, etc.(Feist, Lehmann).

3.1.1.6. naqad- >naked<, Runic (acc. sg.) nAkdan, OE nacod, OFris.nakad, OHG nac(c)kot, nahhut (Lehmann).

3.1.2. Material with weak or no general support in Germanic:First the -odus/-ofws group. The considerable variability in the ac-

cent position in IE words containing the *-tu- suffix rests to some




extent on the Opposition of accented füll grade of the root vs. suffixaccent accompanied by zero grade of the root, an Opposition that isstill clearly preserved in Vedic (Brugmann 1906, p. 440-446). Further,judging by the non-palatalized velars in Ved. gantum : gatva, kartum:krtva, the füll grade probably represents o-grade. This provides someindirect indications of accent position which acquire a particular im-portance in view of the productivity of the compound suffix *-odu-/•ofm- in PGm. (ibid., p. 445).

Krause (1971, p. 40) has suggested that the Variation in the cognatesuffix attested in three Runic words is due to Verner's law. Two ofthese fit the expected accentual distribution exactly, viz. lafjödu >invi-tation<, with root-accented o-grade (Kluge/Seebold 1989, s.v. laden 2),and unapou >contentment<, with zero grade root and accent on thefirst part of the suffix. The third item is haukopu^ with o-grade diph-thong losing accent to the following syllable, rather, apparently, äs inGoth. flodus >stream<, OE (runes) flödu >flood<, OE, OS, OFris. flöd,OHGflöt,fluot (Feist, Lehmann).28

Three of the five Gothic items cited for this suffix seem to fit thissame pattern, viz.:

3.1.2.1. wratodum (dat. pl.) >journey<, with accented o-grade of theroot;

3.1.2.2. gabaurjolwm (dat. pl.) >pleasure<, with zero grade of theroot and accent on the following syllable;

3.1.2.3. *gaunofw (acc.) >mourning, sorrow<, losing accent from theo-grade diphthong to the following syllable. However both manu-scripts have gaunot>a and the emendation has been made on the basis

28 Besideflodiis, Brugmann (1906, p. 441) cites Goth. skildus vs. daupws,wultms äs illustrative of Variation in accent position in PGm. wordscontaining the *-tu- Simplex. Curiously enough, these four items alsoreveal a sort of compatibility with Thurneysen's law, except that herethe determining consonants are the root initials: voiceless ./Z-, sk- vs.voiced rf-, w- (the same is also true of blo]> >blood<, the only one ofSuzuki's [1994, p. 235, fn. 20] eight non-analogical items [Woodhouse1998, p. 195f.] with medial dental), yet all four of Brugmann's examplesdo indeed derive their consonantism from PGm. Thus skildus >shield<agrees with OIcel. skipldr, OHG skilt, apparently with suffix accent,despite e-grade of the root - unless this is another of my >intermediatezercx grades (fn. 4 above); daufms >death< agrees with OE dgafr, OFris.däth, död, OG töd, all with root-accented o-grade; wulfws agrees withRunic wl{yul)ewaR (man's name), OIcel. UUr (god's name), all havingaccented zero grade of the root.




that the emended suffix is proper to abstracto derived from the -class of weak verbs. Such a connection is by no means obligatory,however, cf. faked- >joy< : faginon >rejoice<, frijafnua : frijon >love<.Consequently, it may be preferable to take gaunofxi (acc.) at facevalue and see in it a *-£ abstractum based on the verb gaunon. Ac-cent just before the suffix of such abstracta appears to have beencommon (Brugmann 1906, p. 416-418).

The following additional item can be safely included here:3.1.2.4. manniskodus >human nature<: the only certain information

on accent is supplied by Greek, where the diminutive suffix -ίσκο- isconstantly so accented (Brugmann 1906, p. 501 f.); in Germanic onlyOIcel. oeska (< *junhiskon >youth<, ibid.) is against this, the *h indi-cating accent on the first syllable. On the other hand Greek -τύ- pre-ceded by long vowel retains its accent s before (e. g. άλαωτύς >actof blinding<); consequently there is evidence for accent on every sylla-ble of manniskodus except -sko-.

Several members of the weitwod- >witness< word family have natu-rally developed -d-, s follows.

3.1.2.5. weitwodipa >testimony, etc.< belongs here on account ofthe -d-, not the -/>-, although, s we have seen (3.0.1 above), it is thelatter that enables us to predict the accent position with a high degreeof confidence.

3.1.2.6. weitwodjan >testify, bear witness<, denominative: accenton -ja- (cf. 2.1.2.5 above).

3.1.2.7. weitwodeins >testimony, etc.<: see riqizeins (2.1.2.10 above).3.1.2.8. weitwodei >testimony, etc.<, -m- stem: the accent in Greek

-iv- stems is uniformly on this suffix, and also on the allegedly relatedVedic ac ectival suffix -in-, cf. also kanina- >young< (: Λν. kainln->girl<) (Brugmann 1906, p. 313-315).

The remaining items are more various in their origins.3.1.2.9. witoda (dat. sg.) (and compounds) >law<, differing in its suf-

fix consonant from OHG wizz d >id.<, but agreeing with OE witod-lice >tmly, certainly<. The meaning >law< of the Gothic word derivesreadily from *>that which is/has been observed, adhered to< (hardly**>what has been seen<, pace Lehmann), indicating a passive participlefrom a denominative *witon (Feist) *>practice observance of, adher-ence to< related to primary witan >watch, observe<. The -d- of theGothic word agrees with the characteristic accented suffix *-id- of theparticiple (Brugmann 1906, p. 400), not *-tu- (pace Lehmann), sinceGothic would have preserved the u, s in items 3.1.2.1-4 above.




3.1.2.10. naqadei >nakedness<, Verner's law s for weitwodei(cf.3.1.2.8above).

3.1.2.11. liuhada (dat. sg.) >light< has suffix *e/o-£d-, cf. Ved. vr-at -m>commandment<, Gk. βρ-οτό-ς >mortal< (Brugmann 1906, p. 401); themedial -h- is then due to arrest of Verner's law voicing (see 6.2 belowfor the chronological, etc., questions this raises).

3.1.2.12. liuhadei, liuhadeins, the -a- can be due to Verner's law s in weit-wodei(ns) (cf. 3.1.2.7-8 above); the -h- s for liuhada-, but the consonant-ism of the words may be entirely adopted from liuhada (see preceding).

3.1.2.13. mitade (gen. pl, fern. cons. stem) >measure, amount mea-sured out; device for measuring (?)<, a fern, abstractum, according toBrugmann (1906, p. 426), but the Vedic items with alleged accentedsuffix *-e/o-t- adduced s possibly comparable by Brugmann (sravat-,etc., 1906, p. 425) are regarded by Macdonell (1968, p. 189) - probablycorrectly in view of the absence of any lengthening by Brugmann'slaw of a putative suffixal *o in items such s jagat- >world<, v ghat->sacrificer< - s having zero grade (*-n£-) of the participial suffix with(Vedic only?) substantivizing shift of accent. A number of Greek andone or two Germanic ί-stems cited in this context by Brugmann (1906,p. 425-427) appear to differ from comparable Vedic material in hav-ing accented e-grade of either root or suffix, cf. Gk. άργέτ-, πένητ-,εχητ-, κέλητ-, λέβητ-, Θέμα-, μέλιτ- also χάριτ·, OS metod (= ΟΕ meo-tod), perhaps also (with o-grade root and accented suffix) OHG leitid,sceffid, OE hattet* (= OS helidvs. ON hqlar with a different develop-ment of the suffix and with indeterminate stem final consonant any-way); these, I believe, make likely, if not absolutely certain, originalaccent on the first syllable of mitad-.

3.1.2.14. mitadjon (hapax dat. sg. - not acc., pace Lehmann mitari)>measure, method of measuring (?)<: the uniform accent on the zerograde *-in- suggests that the PIE accent should fall on some part ofthe f ll grade *-iy n- of the same suffix and is therefore in fact pre-served by the Greek forms in -ίων cited by Brugmann (1906, p. 316),even though in principle PIE accent on any preceding syllable wouldalso result in the same Greek accent place; this conclusion receivesindirect and perhaps somewhat slight confirmation from the suffixaccent evident in the shortened form reflected in Gm. *-jan- in Goth.wai-dedja, etc., >evil-doer< (Brugmann 1906, p. 317), assuming thisgoes back to a PIE *-cfehitiyon-.29

29 The time is perhaps ripe for a ttle more assertiveness on my part in




3.1.2.15. fahedai >joy<: despite the lack of a formally exact Gothicverbal quivalent to this nominal formation, its structure neverthelessappears to be exactly paralleled by Gk. οϊκησις (Feist), φόρησις,κόσμησίς (Brugmann 1906, p. 435f.), all pointing to suffix *-eti-s withaccented o-grade of the root, exactly s in the Gothic word.

3.1.2.16. ahtuda >eighth<: niunda >ninth<, taihunda >tenth< (>seventh<is unfortunately not attested) suggest levelling of/d/ in these consecu-tive ordinals (for discussion see Ross/Berns 1992, p. 622-624, 629-632).


3.2.1. awefn >flock of sheep< = ποίμνη (Feist), not, pace Lehmann,**ποίμην (nor **>herd of sheep<?). On the other hand Lehmann'sjudgement -/>- >by Thurneysen's law< is unassailable beside OE eod(e),OHG ewit, ow i.

3.3. Items whose causal relationship with Thurneysen's lawcannot be determined

3.3.1. auhjodus >hubbub<: no secure etymology; while it would benice to be able to Claim >accented o-grade of the root<, the first syllablecan just s easily continue an exclamatory *uh.

3.3.2. bajofwm (dat. pl.) >both<: etymology and structure uncertain;may originate in a compound like OHG beide, bede, etc. (Feist, Leh-mann).

3.3.3. -du{>, i-stem suffix in mikildufxiis (gen.) >greatness' and ga-maindupais (gen.) >community<, etc. That the suffix itself representsa productive levelled form in Gothic is vouched for, unless an in-tervening vowel has been lost, by the clusters -kd- and -gd- (insteadof **-/&£-) in ajukdul) (acc.) >eternity<, managdufts >abundance<. Thesuffix is found only in Italic, Celtic, Iranian and Gothic, not in Indicor Greek (Brugmann 1906, p. 453 f.), so there is no way of checkingits accentual characteristics. The first consonant of the suffix mustbe due to Verner's law (gamainfrs >congregation<, OHG gimeinida

employing the usual Symbols for implosives to denote the PIE conso-nants traditionally regarded s having been voiced aspirates (seeWoodhouse 1995 and 1997).




>community< scuttle any possibility of analogy) and nothing is againstassuming that the accent was on the -u- of the suffix, but there isnothing to prove this either (certainly not the nomen actionis Gk.μνηστνς >an asking in marriage< cited in this connection by Brugmann,ibid.)·

3.3.4. dalapa >below<: a hapax (Mk. 14,66). If the final -a is not adittography (the next word is atiddja and a sign of carelessness atthis point in the text is the omission, according to Streitberg [1950],of an intervening *[;α/ι]), the form appears to be quasi-loc./dat. beside>acc.< dalafr. If this is so, then, given the defectiveness of the >para-digm< and the absence of any other Information s to the age of theform, there is little to choose between the conflicting preferences forthe etymology of the Jb of the suffix indicated by Lehmann (s.v. α#οφ,saraajb), viz. allative *-te, locative *-cfi, ablative *-cfew in the Greekinterrogatives πόσε, πόθί, πόθεν and elsewhere, e. g. τηλόσε, τηλόθι>far away<, τηλόθε(ν) >(from) far away<, except for one thing: if voice-less /> is derived from PIE *cf this would represent the only instancein which an old pre-Verner PGm. voiced obstruent was subject toThurneysen's law (see also Schwyzer 1939, p. 627-629 on the etymol-ogy of the Greek desinences; note that the final consonant in Ved.s(u)mat is unhelpful here since it is a product of neutralization).

3.3.5. magafwi >maidenhood<: one expects *-dei s in weitwodei(cf. 3.1.2.8 above), but /> may come from magafr (q.v. 3.1.1.4 above;also 6.2 below). Thurneysen modification is possible, but questionablein the light of the >exception</r^modei (cf. 3.4.1 below).

3.3.6. weitwode (gen. pl.) >one who witnesses<: concerning the valid-ity of the direct comparison of this with a perfect participle, viz. Gk.είόώς, εΐόότος, despite comparison of Goth. berusjos, f^ewisa with thesame member of the verbal paradigm, see Brugmann's (1906, p. 565 f.)Suggestion of PIE *-wot- beside *-wes- (cf. also Bammesberger 1986 a,p. 103). As has perhaps been hinted at above (2.1.2.9), the e-grade inthe first syllable means that Gk. εΐόότ- with accented o cannot beoriginal. Apart from this source of uncertainty, there is also fairlyheavy analogical pressure from the rest of the family (3.1.2.5ff.) infavour of the -d-.

3.3.7. weitwodi >testimony, etc.<: accent of *-(i)t/o-stems appears tobe very variable judging by the examples collected by Brugmann(1906, p. 182-193) and there is the possibility of analogical interfer-ence (see preceding 3.3.6).





3.4.1. frumadei >pre-eminence<, with the voiced -d- expected withthe -m- suffix (weitwodei, cf. 3.1.2.8 above).

3.4.2. arbaidim, (dat. pl.) >work, labour, toil, trouble, etc.<: agreeswith OHG ar(a)beit, OFris. arbed against OE earfoh OS arbed, etc.(Feist, Lehmann).

3.4.3. arbaidjan >work<: the -d- is expected in a denominative; thedifferences between OHG arapaitön >work< and OE earfofnan >bur-den< (Feist, Lehmann) point to separate levelling processes within thelanguages (cf. arbaidim).

3.4.4. haubida (dat. sg.) >head<: -d- well supported in Germanic, cf.OS höbid, OHG houbit, etc. (Beekes 1996, p. 218).

4. Data for alternating velars -h-l-g-

Ignoring the enclitic -uh with its various peculiarities, and also-hun, äs being beyond useful analysis here, the material reduces togroups containing the following >suffixes<: -ah/-ag, -ih/-ig, -eig and-ug, together with some items containing further derivational Suffixes,especially -ei(n-).

In a special study of nearly all of this material, Bammesberger(1986 b) challenged the communis opinio that the alternants here arepurely the result of Verner's law (p. 35) by raising doubts that thiscould be so for any of the material containing the voiceless alternant/h/ (p. 36-38). Bammesberger concluded that since, äs Thurneysen(1898, p. 212) himself saw, all /i-forms are in harmony with Thurn-eysen's law, Thurneysen's original formulation must have been cor-rect, i. e. that the alternants were originally phonologically determinedby his law and that the large number of exceptional cases involvingthe voiced alternant /g/ were the result of subsequent levelling (1986b,p. 37f.).

While Bammesberger's separate conclusions regarding the indivi-dual fe-items can in general be upheld, his overall conclusion attribut-ing the non-conforming 0-items to levelling is another unhappy childof the piecemeal approach that has characterized this subject to date.The uniformity in the direction of the alleged levelling, which is stillsomewhat worrying even when viewed in Isolation, becomes quiteincredible when placed alongside the forms containing -ifxi-. As Bam-mesberger (1986 b, p. 35) himself points out (somewhat cursorily), the




available comparative evidence for the /i/0-items indicates that theSuffixes in question overwhelmingly continue Proto-Germanic */g/arising from Verner's law, since the accent in comparable material inVedic and Greek is almost never on the vowel preceding the */c.30

The -ipa- case is essentially the mirror image of this (cf. 3.0.1 above).The levelling theory thus requires that in two separate cases the origi-nal preponderance of one of the alternants dictated by Verner's lawwas first obliterated by Thurneysen's law and then the same prepon-derance was restored by analogy in the same unidirectional way inboth cases, and, be it noticed, in the absence of any competing casein which analogy might have favoured an originally non-preponderantalternant. This is surely pushing coincidence too far. It is surely muchmore likely that the preponderance of the (j-forms in Gothic is a directreflection of the Situation obtaining in Proto-Germanic immediatelyposterior to Verner's law.

Although this conclusion implies a very different view of Thurn-eysen's law than that espoused by Bammesberger, its immediate prac-tical effect for present purposes is about the same, viz. it is only theft-forms that need be subjected to detailed analysis of their Thurn-eysen Status. Much of this task, s has been indicated, has alreadybeen done for us by Bammesberger (1986b), although in some in-stances in a curiously non-explicit way - see, e.g., his remarks on

30 Exceptions generally have long vowels before the suffixed k, viz.zandtka- >egg-laying< (Brugmann 1906, p. 496), parphartka- >filler<(Macdonell 1968, p. 118), s ctka-s >stinging insect< (Lehmann bairga-hei), asmaka- >our< (Feist ainoho, Lehmann ainaha). But cf., with ac-cented short vowel, Gk. πηλίκος >how much? (in size, quantity, age)<,etc. (Brugmann 1906, p. 496). - For -ag-/-ah- cf. Ved. ekaka- >only<,asvaka- >horse (dimin.)<, sanaka- >old<, duraka- >distant<, arbhaka->small< and nagnaka- >naked<, mamaka- (and mamaka-) >my<, antaka->ending, destroying<, rupaka- >adopting a form<, sayaka- >intended forthrowing< (Brugmann 1906, p. 493). - For -ig-l-ih- cf. Ved. pary yikd->strophic<, varsika- >of the wet season<, Gk. Ιππικός >horse's<, νυμφικός>bride's< and many others (Brugmann 1906, p. 488). - For -eig- cf. Ved.(mostly subst.) mrdika- >gracious; grace<, anika- >face<, dfsika- >as-pect<, asarlka- >rheumatic pains<, dfbhtka- (N. of demon), vrdhika->increaser<, etc. (MacdoneU 1968, p. 118; Brugmann 1906, p. 495 f.). -For -ug in handugs cf. Ved. tanuka- >thin<, sisuka- >little child<, s n-uka- >desirous of prey<, etc., Gk. Αιβνκός >Libyan<, Θηλυκός >feminine<(Brugmann 1906, p. 491; Macdonell 1968, p. 120), remembering alsothat Goth. handus >hand< is a w-stem. - For the suffix -ei(n-) cf. weit-wodei (3.1.2.6 above).




bairgahei, unbarnahs, broprahans (1986b, p. 36f.). It remains onlyto present the results, together with any additional commentary thatmay appear necessary, in a manner compatible with the remainder ofthis survey.


In principle this simply includes all the 0-fonns that ostensibly com-ply with Thurneysen's law; the remaining ^r-forms will, äs indicated,also constitute Output from Verner's law but will be exceptions toThurneysen's. In the case of the -eig- suffix the numbers in each ofthese groups are roughly equal (eight vs. nine). Of all the remaining(j-forms only vwlfrags belongs to the conforming group; the rest - atotal of some 19 items, counting derivatives - count äs >exceptions<.31

Under these circumstances further division of this material into thosewith and those without Germanic support outside Gothic seemsrather pointless.

As might be expected, none of the /i-forms can be shown with anyreasonable degree of probability to have contributed to the genesis ofthe law. For some items there are nevertheless certain slender hints.These items are therefore presented in a special group under the >un-certains< (cf. 4.3.1 below).


4.2.1. aurahjom (dat. pl.) >tombs<: cf. mitadjon (3.1.2.14 above); inaddition derivation via, among other things, a collective based on*4n- (Feist and, with the addition of OIcel. aurigr, Lehmann) leadsto expectation of **-#- (unless a connection with OHG -ahi collectivescan be made good, see 4.2.6 below).

4.2.2. broprahans (acc. pl.): the other Germanic evidence (Feist,Lehmann) attests solid /g/.

31 From Thurneysen's original list of 19 such items, one, »wainags«, is tobe subtracted, since it is now read either wainahs or wainans (Feist,Lehmann; Bammesberger 1986 b, p. 37), and a new one, hailag (neut.)from the Pietroassa ring, in all probability to be added (Bammesberger1986 b, p. 35, fn. 1; cf. also more recently Nedoma 1993 b and Reichert1993).




4.2.3. unbarnahs >childless<: no exact Germanic cognates for thiswordform (Feist, Lehmann), which undoubtedly represents a pro-ductive type (cf. e.g., OIcel. ublodigr >unbloody<, which shows theexpected general Germanic form with /g/ but is otherwise uninforma-tive).

4.2.4. waurdahs >characterized by words (?)<: hapax in context (Sk.4,18) that makes determination of precise meaning extremely difficult(some sort of compromise between >heavenly< and >earthly<?); Ger-manic cognates (meaning roughly >characterized by many words<)support only /g/ (Feist, Lehmann; Bammesberger 1986 b, p. 36).

4.2.5. wainahs >miserable, wretched<: hapax; reading with -h- possi-bly incorrect; Germanic cognates support only /g/ (Feist, Lehmann;Bammesberger 1986b, p. 37).

4.2.6. stainahamma (dat.; acy. functioning s subst, perhaps in imi-tation of the Greek) >(the) stony (i.e. ground)<. AU the other Gm.>stony< words, except OHG steinahi >stones, stony ground<, attestsolid -g- (Feist, Lehmann). Several arguments against citing OHG stei-nahi in support of -h- in the Gothic word are presented by Bammes-berger (1986b, p. 38). There is another argument in the same vein thatis worth mentioning: since there is a well-attested group of OHG -ahicollectives traditionally regarded s containing the collective suffix*-ja- (Thomas Klein [Bonn], personal communication), it is also pos-sible that the OHG suffix -ahi represents a (possibly generalized) PIE*-ok(i)yo- with accent position determined not by the *-ko- suffix but,s might be expected, by the *-(i)i/o- suffix, since Gk. θειος, θείος

>divine< : θεός >god<; ήμιος >honoured, honourable< : τίμή >honour<;άγριος >w d < : αγρός >field< (beside Skt. ajra- >plain, meadow<; Brug-mann 1906, p. 188 f.) suggest that the *-(i)yo- suffix can send the ac-cent back to the syllable before itself. Note that the Greek femininecollectives in -ία (φρατρία, etc.; Brugmann 1906, p. 647) follow Greekaccentual principles and are non-probative for PIE accent - cf. ξενία>hospitality< beside probative accent in ξένιος >hospitable< (Brugmann1906, p. 189). The question of whether the OHG neut. collectives in-ahi have influenced the consonantism of Gothic fern, collectives in-ahein- (see aurahjom 4.2.1 above and bairgahei 4.3.1.2 below) canbe reserved for another occasion: it may have a small but non-signifi-cant effect on the results of the present investigation.




4.3. Items whose causal relationship \vith Thurneysen's lawcannot be determined

In view of the dearth of >velar< material that could definitely haveserved s the basis for Thurneysen's law it seems appropriate here to sep-arate items for which there is even the slightest indication that they couldhave fulfilled this function from the completely uncertain material.

4.3.1. Material with faint indications of possible inherited or >inde-pendent< Status:

4.3.1.1. ainaha (masc., 2 x), ainoho (fern., hapax, usually held to bea scribal error for ainaho) (all in Luke) >μονογενής, only (begotten)<.Germanic cognates are solidly in favour of PGm. /-g-/ (Feist, Lehmann;Bammesberger 1986 b, p. 36); but in view of (1) the paucity of attesta-tion, (2) the etymological connection - fortuitous though this maybe - between Gk. -γεν- in μονογενής and Eng. kind used in glossingsome items containing the suffix found in Ved. asm ka- >our<, RUSS.odinakij (obsol.), odinakovyj >same, identical, of one kind<, dvojdkij>of two kinds< beside odinokij >solitary< (Feist, Lehmann, citing Kluge)and (3) the ever present possibility of confusion between words forthe concepts >solitary<, >unique<, >one of a kind<, >only begotten< and>of one kind<, it must be said that the scribal error theory for ainoho,though persuasive, is not quite conclusive. Instead it is just possiblethat Gothic preserves the vestiges of a competition between thesetwo similar forms, with the (voiceless) consonant of the >variant< withlong accented vowel (cf. m. 30 above) levelled into the other forms.But this is, admittedly, a long shot, and besides, it would hardly bepossible to substantiate any claim that the putative leveliing had pro-ceeded independently of Thurneysen's law.

4.3.1.2. bairgahei >mountainous region<, with -ei(n-) suffix. Againstaccepting Thurneysen's law s the sole explanation is the possibilityof/g - g/ dissimilation. Cf. also the case ofmagafrei (3.3.5 above); andalso OHG steinahi (4.2.6 above). Feist, Lehmann make no attempt todiscuss the consonant of the suffix: both refer the reader to stainahs,where however the reader is referred straight back to bairgahei', bothauthors refer to Kluge's >Stammbildungslehre< of 1926, but both limitthemselves to an expression of disagreement with the latter's detailedtreatment of ainoho. While their disagreement is understandable, itmight have merited a little more discussion since ainoho surely repre-sents the most important (even if ultimately erroneous) piece of evi-dence for (pre-Thurneysen) PGm. */h/ in these Suffixes.




4.3.1.3. frarihis (gen.) >unshrunk<: hapax; »Deutung [...] völlig un-sicher« (Bammesberger 1986 b, p. 37); no certain Germanic cognates(Lehmann; Feist's emendation and comparison with OHG derh >holed,holey< would remove the word from consideration here). Eb-binghaus's cogent emendation to *pareihis (Lehmaim) opens somepossibility of PIE accent on the long vowel (cf. fh. 30 above), whichwould supply a needed example with post-Verner voiceless velar andalso transfer the >g-oiuy< Status from the -eig to the -ig group.

4.3.2. Completely uncertain material:4.3.2.1. niuklahs >child, immature person<: no judgement possible in

view of uncertain etymology (Feist, Lehmann; Bammesberger 1986 b,p. 37).

4.3.2.2. niuklahein (dat.) >pusillanimity<: äs preceding, which mayhowever have supplied the /h/.

4.4. Exceptions to Thurneysen's lawThese have been discussed in a general way in section 4.1 above. I

see little point in listing the relatively large number of forms individu-ally. The precise number to be considered is not really very importantexcept insofar äs it may affect perceptions regarding the validity ofthe levelling theory espoused by Bammesberger (1986 b). Eb-binghaus's emendation of parihis (cf. 4.3.1.3 above) additionallyprompte caution in evaluating the nine >exceptions< among the other-wise apparently non-relevant -eig group. The remaining 19 excep-tional items already referred to comprise: 13 -ag-, 3 -ig-, l -Heg-, 2-ug- (see Thurneysen 1898, p. 212; and fn. 30 above).

5. Data for alternating labials -f-/-b-

The small amount of well-known material to be discussed here re-quires at most a Separation of the exceptions from the items thatconform.

5.1. Material conforming to Thurneysen's lawThere is no material inherited with the Thurneysen Variation al-

ready in place.The five well-known items belonging to the -ubnil-ufni group

clearly have their suffix consonantism äs a result of Thurneysen's law.




These preserve the zero grade of the suffix *-men- occasioned by thefurther addition of the suffix *-y-o/ä- (Feist, Lehmann). In every casethe grade of the root in the Gothic words simply duplicates that ofthe ostensible Gothic derivational base (both being /o/ or both zero,in cases where the grade is known) and has no effect on the conso-nantism of the suffix,32 which is also independent of the gender of thederivatives, and thus of their presumable original semantic bifurcation(possibly into fern, of result vs. neut. of action).33 Thus without evenbothering to consider the question of whether the *ra in items likethese could be affected by Verner's law, it is clear that the only pos-sible explanation for the consonantism of the suffixes in these wordsis Thurneysen's law.

One item - daufmbljans >sentenced to death< - with no certainetymology and every prospect either of being a compound34 or of

32 Flickinger's (1981, p. 71) forlorn comparison of witubni with Ved.vidman- ignores not only the difference in word structure, but alsothe levelling of grades and/or accent position in the paradigms in bothlanguages, and even the agreement between the suffix accent and thegrades of root and suffix in the two recorded forms (inst. & dat. sg.)of this particular Vedic word (cf. Beekes 1995, p. 186; Macdonell 1968,p. 207-210).

33 These facts are conveniently summarized in the table:

Derivative Grade Baseof root

Form

fraistubnifastubni

witubniwunduftüwaldufhi

Gender

fern,neut.

neut.fern,neut.

Meaning

>temptation<>observing(rules); fasting<>knowledge<>plague, illness<>power, authority<

9

/o/

zerozero/o/

Form

*fraistanf äs tan

witanwundswaldan

Meaning

>tempt<>observe;fast<>know<>wounded<>rule<

For *fraistan Gothic has/raisan, reflecting the commoner protoform.Lehmann suggests a compound/ra-is-; Mann (1984-1987, col. 987 f.) ap-pears to hesitate between this and a compound based on a PIE prefix*prdi--, but in view of the semantic parallel in NHG locken <attract, tempt<beside Locke >lock of hair< (Kluge/Seebold 1989 s. w), it may also beworth considering the possibility of an ablaut-revealing connection offraisan with OFris. frlsle (also/reste) >(lock of) hair<, OE/rts >curly<.

34 Curiously, perhaps because the development *ml- > bl- is not usuallyinvoked for Germanic (but cf. Kluge/Seebold 1989 s.v. Blei 1; also theseveral Germanic bl-words for >white(n), pale, flash, mildew<, etc., be-side Mir. mlicht, blicht >milk<, etc.), the etymologies that have been




simply complying in every particular with Verner's law, belongs hereto the >indeterminate< group.


The only items of any interest here35 are the Gothic numerals >11<and >12<, both of which have -lib- (PIE *4ip- or *4ikw-\ contrastingwith OHG -lif- for both, äs Thurneysen (1898, p. 212) points out, add-ing however the incorrect inference that the two OHG forms mustcontinue the original PGm. ones. Clearly both forms of this suffixmust have been present in PGm. immediately after the Operation ofVerner's law (cf. Ross/Berns 1992, p. 594-597), and no doubt eacharose originally in only one of these numerals, the later distributionbeing due to levelling between consecutive numerals within the sepa-rate languages. In Gothic, equally clearly, this has proceeded in sub-lime disregard of Thurneysen's law.

6. Summary and conclusions

6.1. The origin of Thurneysen's law

The distribution of the items investigated above can best be sum-marized in tabular form. The material falls naturally into two groupsaccording to whether the alternating suffix consonant is apical (dentalor Sibilant, table 1) or non-apical (velar or labial, table 2).

Not included in table l are the Jb/d items of 3.0-3.0.2 since theargument to be presented below emerges sufficiently clearly, I be-lieve, without the help of Üus not wholly relevant material (see table3 and associated discussion below).36 The counting of the items has

proposed to date seem not to include an appeal to the zero grade ofPIE *mel- >speak< äs in NHG melden >report< (Kluge/Seebold 1989 s.v.),zero grade in RUSS, molvit? >say< (Vasmer/Trubaöev 1986-1987 s.v.); cf.also Russ. kuda meles? (lit.) >which direction are you grinding in?< =>what are you implying?< = >what are you saying?<.

35 The complete immunity of the adverbial suffix -ba from Thurneysen'slaw is probably due to recent origin in Gothic by way of a compound-ing element and so provides no Information on the susceptibility orotherwise of PIE asperae to this law (cf. Heidermanns 1996, p. 273).

36 For similar reasons of fairness to alternative views, and also in viewof the uncertainty raised by Ebbinghaus's emendation for {xirihis(cf. 4.3.1.3 above), the -eig items (eight conforming vs. nine excep-tions) have been included in the figures for table 2.




not been entirely consistent: e.g. weitwodei and weitwodeins havebeen counted separately, liuhadei(ns) äs a single item, but this hasprobably helped to assign some sort of weighting to variants of thistype (which are actually homographic in some forms, e.g. acc. sg.). Amore thorough study might seek to establish the relative frequenciesof the various forms in speech; but since the ecclesiastical texts avail-able to us can hardly count äs samples of daily Gothic chitchat, it isvery questionable whether such a labour would be rewarded by anyreal increase in the reliability of our conclusions. Many other detailsof the material are no doubt open to criticism and revision, but Ibelieve the general picture represented by the above figures is reli-able, despite the overall smallness of the figures for the purposes ofstatistical computation.

Table 1: Distribution of Thurneysen-relevant material with apical suffixconsonants

Altern. Material conforming to Thurneysen's law Exceptions tosuffix Thurneysen'sconson. law

Due to Due to Indeter-Verner Thurneysen minate

s 5 l 5 -z 7 l 1 2b 7 l 4 -d 1 5 - 3 4

Total 34 3 13 6

Table 2: Distribution of Thurneysen-relevant material with non-apical suf-fix consonants

Altern. Material conforming to Thurneysen's law Exceptions tosuffix Thurneysen'sconson. law

Due to Due to Indeter-Verner Thurneysen minate

h - 5/6? 5g 9 - 28f - 2 -b - 3 l? 2




Within the two apical (sibilant and dental) categories, despite theabove provisos, I think there is a clear preponderance of material thathas in all probability acquired the Thurneysen Variation äs a directresult of Verner's law. Even in the worst case scenario in which 100%of the >indeterminate< items had their suffix consonantism determinedby Thurneysen's law, the Situation immediately post-Verner is that, ofthe attested items, 34 already conformed to the Thurneysen patternagainst 22 (= 3 + 13 + 6, the totals in the remaining three columns)that did not, or a ratio of better than 3:2.

Consequently, I believe that these figures support the view that de-rivatives containing apical Suffixes ostensibly conforming to Thurn-eysen's law were inherited by Gothic from the Operation of Verner'slaw in sufficient quantity to engender a feeling among Speakers thatthis inherited >Verner/Thurneysen variation< supplied a criterion forwell-formedness in suffixal derivation.37 Consequentiy this inheritedphonetic Variation began to spread to other similar derivatives that didnot originally have it, first to those with Suffixes based on apical spi-rants and later to those with Suffixes based on the two peripheral (labialand velar) spirant categories. Neither of the latter two categories werelikely to have given rise to such a development themselves since *-p-suffixes were rare to non-existent in PIE (Brugmann 1906, p. 386, fh.1) and *-A>suffixes, äs we have seen, seem almost invariably to havehad the PIE accent in positions not conducive to the development ofGermanic forms alternating on the basis of voice. The difference inthe success rates of this spread among labials vs. velars seems to beidentifiable with the much larger quantity of non-conforming Verner-determined material in the velar group, although it is also possiblethat [+anterior] represented a favoured feature in this context.38

The figures demonstrating the preponderance of material withVerner-outputted Thurneysen Variation could certainly be improvedby including the material of 3.0 rejected äs unaffected by Thurn-eysen's law since this material more than doubles the Verner-out-putted conforming total to 70 vs. a worst case increase of the immedi-ately post-Verner non-conforming items to 32 (= 3 + 13 + 16, see table

37 The question of well-formedness of suffixes can be readily illustratedfrom the distribution of NHG -heiti-keit (Wright 1907, p. 151 f.); cf. alsothe replacement of NHG -bar with -lieh for stems ending in b (e.g.unhaltbar vs. unglaublich), and the like.

38 Very likely it was due to both these factors acting in concert, togetherno doubt with other factors that we can hardly even guess at.




3), yielding a ratio of better than 2:1. The figures have been obtainedby adding to Thurneysen's (1898, p. 211) 25 conforming examples (vs.six exceptions) with -ipa and two with -ida, s well s his (ibid.) threedistinct conformers with -pwa (vs. one exception with -dw) and fiveconformers (vs. two exceptions) with -pro, our newly discovered *ful-lifra (cf. 3.0.1 above) and the hitherto disregarded >exception< gaidw.Quite possibly table 3 represents a truer picture of the Situation thatset the Thurneysen restructuring in motion. Still better approxi-mations to the truth could be obtained by doing a comparative lexico-statistical analysis of material with these suffixes in other early Ger-manic dialects in order to form an estimate of what proportion of thetotal may have been preserved in the Gothic texts.Table 3: Distribution of Thurneysen-relevant material with apical suffixconsonants augmented by non-Thurneysen items of 3.0 f.

Suffixdetails

Table 1-φ/da-φ/dw-φΓ-

Material conforming to Thurneysen's law Exceptions toThurneysen'slaw

Due toVerner

342835

Due toThurneysen

3

_

Indeter-minate

13 6622

Total 70 3 13 16

Some points of interest in the figures - not all of which, admittedly,are likely to stand up to rigorous statistical analysis - are:

(1) Exceptions to Thurneysen's law (including even fnwadw, cf.3.0.2 above) all involve >unexpected< voiced consonants. Only the -ipamaterial departs from this principle, which is perhaps another reasonfor not regarding it s belonging to the Thurneysen corpus.

(2) The preponderance of voiced consonants among the Verner-out-put suffix material conforming to Thurneysen's law is to be expected,since the material necessarily descends from protoforms with a mini-mum of three syllables. Since a voiceless spirant is generated in asuffix by accent on only one of the three, etc., syllables of the proto-form and a voiced spirant by accent on any of the other two, etc.,syllables, we can in general expect the Verner-output material to con-tain at least two voiced spirant suffixes to every voiceless spirant one.




(3) Despite the preponderance of voiced consonants among the>Verner-output< apical material, and the small figures involved, thereis a relatively good balance between the voiced and voiceless al-ternants in this group. Without this balance there would have been noThurneysen's law.

(4) It has been shown that the spread of the Thurneysen distribu-tion was achieved by both voicing and devoicing of suffix consonants,just s Suzuki (1992) claimed. Actually it is only devoicing - whichwas a particular focus of Suzuki's attention - that has been shown tooperate at all places of articulation. However, not too much weightshould be placed on this, since it is likely that some of the gaps inthe picture would be filled in if more was known about the doubtful>indeterminate< material.

6.2. The chronology of Thurneysen's law:implications for Verner's law in Gothic

The proposal that a few items, viz. liuhada, liuhadei and artv-aznoSy may have their determining (i. e. root final) voiceless conso-nant due to the failure of Verner voicing furnishes confirmation thatthe process involved here was not a reversal or elimination of voicingbut simply the arrest of the Implementation of voicing.

In order to demonstrate this, let us begin by assuming the opposite,i. e. that the peculiarities of Verner in Gothic were indeed due to areversal of voicing. Now if items like liuhada, liuhadei and artv-aznos, contributed to the genesis of Thurneysen's law, then Thurn-eysen's iaw must have come into being later than the reversal. If thatis so, then it is a little odd that frewisa and magah with their oldvoiceless spirants, retained their Verner voicing, apparently for noparticular reason, and thus came by Chance to comply so neatly withthe Thurneysen alternation when it eventually came to the language.On the other hand if Thurneysen's law preceded the reversal, wewould have to suppose that it mysteriously failed to apply to the liu-had-group, so that when the reversal set in it simply happened to pullthese few items neatly into the Thurneysen line, without, however,affecting the maga]? group. For the latter, it does not seem possibleto claim Inhibition of the reversal due to the voiceless spirant alreadyoccupying the following syllable, because this would be contradictedby hauhijxi and weihifxi (Thurneysen 1898, p. 211), unless of coursewe acquiesce in the levelling theory for the -i/κι suffix after all.




There are several possible approaches to this little teaser, assumingwe wish to retain the notion of a reversal rather than an arrest ofVerner voicing in Gothic. One is to dismiss it on the account of thesmall number of cases, pointing to the two -i/xi->exceptions< just men-tioned, which lose their solitary, erstwhile putatively voiced spirant,s well s other exceptions, such s ubizwa and tnwadw, where both

voiced spirants are retained.Another approach is simply to conclude that the liuhad-group must

have all had root accent after all. This seems to be quite plausible in thecase of arfaaznos, s we have seen. For liuhada it would be possiblebut not particularly advisable to point to the somewhat uncertain evi-dence (cited by Brugmann 1906, p. 401) of Gk. μητίετα, άριδείκετος(but this probably has the accent typical for a compound) and(Boeot.) καιέτα (which may have advanced its accent frorn the firstsyllable by the usual Greek rules if the long final vowel is genuine).

These difficulties are easily resolved under the proposal that thevoicing of lenited spirants generated by Verner's law was simply ar-rested,39 s explained in my previous paper (1998).40

30 They are also resolved by Vbyles's early East Germanic accent retrac-tion, which is also alluded to in my 1998 paper (p. 197). However, myhypothesis possesses two distinct advantages by comparison with thatof Voyles, viz. (1) in my hypothesis the critical process affecting Gothic,i. e. the arrest or abandonment of the Verner voicing process, can betied to a historical or archeological event, viz. the 3rd Century Gothicmigration; (2) my hypothesis of Verner lenition applying to all cate-gories of Germanic consonants, when combined with a slight changeof viewpoint regarding Normier's (1977) hypothesis of a Germanic Ne-benton, provides, for the first time, an intra-Germanic linguistic motiva-tion for the Germanic accent stuft, s follows. The Verner lenitioncaused non-lenited consonants to become signposts of accent on theimmediately following syllabic nucleus in two environments: (a) imme-diately before Normier's Nebenton and (b) word/root initially. Thus thenon-lenition that began s a f nction of word/root initial position be-came reinterpreted s a sign of word accent, whereupon any later ac-cent in the same word was eventually given up. In ON vedr Kortlandt(1978c, p. 112, and 1988, p. 4) has in fact unwittingly adduced an exam-ple of a post-Verner non-lenited aspera (cf. Schwyzer 1939, p. 533, andWackernagel/Debrunner 1954, p. 725, for the commonness of Ist sylla-ble accent in nouns containing this suffix in Greek - especially neu-ters - and Vedic). Some evidence supporting my wider view of theVerner lenition I hope to present in due course.

40 Responses firom colleagues indicate that two key features of my con-ception of Verner's law may not have emerged s clearly from my 1998




Indeed, äs is also suggested in my 1998 paper, the persistence ofuniquely Gothic items like liuhada may have acted äs the trigger forThurneysen's law. Possibly against this, however, is the Suggestion

paper äs I would have wished. These are (1) that the complete Imple-mentation of the voicing that resulted from the Verner's law lenitionextended over a considerable period and was probably completed inGermanic generally after the 3rd Century Gothic migration; and (2) thatthe lenition of the PIE consonants I refer to äs asperae, which resultedin voiced spirants, was a feature of Verner's law, not, äs some wouldhave it, Grimm's and that Grinun's law äs it applied to the non-lenitedallophones of these consonants - viz. the loss of the PIE feature ofimplosion - took place in some dialects after the completion of thevoicing by Verner's law. However, the phonetic incongruity of the pro-posal that a series of murmurs should be more strongly voiced than aseries of plain voiced stops, together with other factors discussed atlength in my 1997 and 1998 papers, dissuades me from adopting thenotion that has been entertained in some circles that Verner's lawessentially applied voicing to a series of aspirated stops (Bernard Mees,pers. comm. July 1998, p. 7). Kluge's law, which I linderstand has beenadvanced in support of this view (Bernard Mees, pers. comm. Oct.1998, p. 2), represents in all probability a syllable-final merger of allthree PIE stop series in PIE/pre-Gm. mediae in the specific environ-ment before (consonantal) nasal and therefore, despite seductive ap-pearances to the contrary, has nothing in common with Verner's law,which was a process of lenition. Verner and Grimm are thus inter-twined in my conception according to the following conventionalizedchronological scheme (D' = implosive, D = plain voiced stop, T = plainvoiceless stop, Th = voiceless aspirated stop, P = voiceless spirant, D =voiced spirant, D = lenis T, D = lenis >; the dialectal split in stages Vüand VIII is on the basis of stop vs. spirant voiced apical with eventualconcomitant [a] = alveolarity vs. [d] = dentality of apical stops):

Stage Asperae Mediae TenuesIIlaüb

IVVVIVII

vra

D'D'D'D'D'/DD'/DD'/D

DDD/T?T// ?

T

TThThf>P/DP/D/DP/D

DVDYD [a] T

D/D/D [d] T(dialectal split)D/D/D [a] TD/D/D [d] T

P/D/D

P/D/DP/D/D

EventPIEGrimm begins

Verner beginsGothic merger lapsesVerner completedreshuffle of voicedobstruents (Grimmdialectally completed)

Grimm finally completed(with loss of implosion)




implicit in the solidly voiced profile of the exceptions to Thurneysen'slaw to the effect that once the lenited spirants had become voiced,they became immune to the law, since this carries with it the furtherimplication that the law came into being originally äs an alternativeor >corrective< principle for assigning lenition/nonlenition rather thanvoicing itself (this is despite the evident end result alluded to in 6.1[4] above). In this form Thurneysen's law (or at least its beginnings)would thus have predated the general lapse of the process of voicingof the lenited spirants in Gothic, an interpretation that can certainlybe applied to the instances catalogued above under 2.2, 3.2 and 4.2 äscomplying with Thurneysen's law in defiance of Verner's.

The material for the labials need not be an argument against thisdespite the relatively common assumption (see Suzuki 1992, p. 34;Woodhouse 1998, p. 204) that the obstruent that arose from the *ra ofthe compound suffix was ab initio voiced. After all, the dissimilationthat gave rise to this obstruent under unknown conditions of accentu-ation may have gone beyond nasality and included also the voicingfeature, especially in any case involving a (putative) Verner or Thurn-eysen nonleniting environment.

REFERENCES

Bammesberger, A. 1965: Old English gycer and Gothic jukuzi, Lg. 41,p. 416-419.

- 1980: Gotisch walisa*, PBB 102, p. 1-4.1986 a: Der Aufbau des germanischen Verbalsystems, Heidelberg.

- 1986 b: Gotische AcUektivbildungen auf *-/wi-, PBB 108, p. 35-39.1995: Zur Vorgeschichte von gotisch berusjos und fadrein, PBB 117,p. 1-8.

Bauer, W. 1958: Griechisch-deutsches Wörterbuch zu den Schriften desNeuen Testaments und der übrigen urchristlichen Literatur, Berlin.

Beekes, R. S. P. 1995: Comparative Indo-European linguistics: an introduc-tion, tr. by UvA Vertalers/Paul Gabriner, Amsterdam, Philadelphia,

- 1996: Ancient European loanwords, HS 109, p. 215-236.Braune, W., and W. Mitzka 1963: Althochdeutsche Grammatik, llth ed.,

Tübingen.Brugmann, K. 1906: Grundriss der vergleichenden Grammatik der indoger-

manischen Sprachen .1, Straßburg.Falk, H. S., and A. Torp 1960: Norwegisch-dänisches etymologisches

Wörterbuch. 2nd ed., Oslo, Bergen, Heidelberg.Feist, S. 1939: Vergleichendes etymologisches Wörterbuch der gotischen

Sprache. 3rd ed. Leiden.




Flickinger, D. P. 1981: Dissimilation in Gothic without Thurneysen's law,Chicago Linguistic Society 17, p. 67-75.

Fraenkel, E. 1962-1965: Litauisches etymologisches Wörterbuch. 2 vols.,Heidelberg, Göttingen.

Friedrichsen, G. W. S. 1939: The Gothic Version of the Epistles: a study ofits style and textual history, London.

Graff, E. G. 1963: Althochdeutscher Sprachschatz, Hildesheim.Grosse, R. 1971: Althochdeutsches Wörterbuch III: E und F, Berlin.Hamp, E. P. 1973: Solutions and problems from Speyer, IF 78, p. 141-143.Heidermanns, F. 1996: Der Ursprung der gotischen Adverbien auf -ba, HS

109, p. 257-275.Holthausen, F. 1934: Altenglisches etymologisches Wörterbuch, Heidel-

berg.Illich-Svitych, V. M. 1979: Nominal accentuation in Baltic and Slavic, tr. by

R. A. Leed and R. F. Feldstein, Cambridge/Mass., London.Kluge, F., and E. Seebold 1989: Etymologisches Wörterbuch der deutschen

Sprache. 22nd ed., Berlin, New York.Kortlandt, F. H. H. 1975: Slavic accentuation, Lisse/NL.- 1977: Historical laws of Baltic accentuation, Baltistica 13 (2), p. 319-

329.1978a: L-E. palatovelars before resonants in Balto-Slavic, in: Recentdevelopments in historical phonology, ed. by J. Fisiak, The Hague,Paris, New York (Trends in Linguistics Studies and Monographs 4),p. 237-243.

- 1978b: On the history of Slavic accentuation, KZ 92, p. 269-281.- 1978c: Proto-Indo-European obstruents, IF 83, p. 107-118.- 1988: Proto-Germanic obstruents, ABäG 27, p. 3-10.- 1994: From Proto-lndo-European to Slavic, JIES 22, p. 91-112.Krause, W 1971: Die Sprache der urnordischen Runeninschriften, Heidel-

berg.Kruopas, J., et. al. (eds.) 1972: Dabartines lietuviy kalbos iodynas, Vilnius.Kurschat, A. 1968-1973: Litauisch-deutsches Wörterbuch. Thesaurus lin-

guae ütuanicae, ed. by W. Wissmann and E. Hofmann w. assist. of A.Kurschat and H. Krick. 4 vols., Göttingen.

Kurschat, F. 1883: Litauisch-deutsches Wörterbuch, Haue.Lehmann, W. P. 1986: A Gothic etymological dictionary based on the third

edition of Vergleichendes Wörterbuch der gotischen Sprache by Sig-mund Feist, Leiden.

Leumann, M. 1977: Lateinische Laut- und Formenlehre, München.Macdonell, A. A. 1968: Vedic grammar, Delhi.Mann, S. E. 1984-1987: An Indo-European comparative dictionary, Ham-

burg.Mayrhofer, M. 1986-1996: Etymologisches Wörterbuch des Altindo-

arischen. 1. Teil: Ältere Sprache. 2 vols., Heidelberg.Nedoma, R. 1993 a: Zur Runeninschrift auf der Urne A. 11/251 von Loveden

HUI, Sprache 35, 1991-1993, p. 115-124.- 1993 b: Abbildungen und NachbUdung des unbeschädigten Runenrings

von Pietroassa, Sprache 35, 1991-1993, p. 226-234.




Niedermann, M., A. Senn, F. Bender and A. Salys 1932-1968: Wörterbuchder litauischen Schriftsprache. 5 vols., Heidelberg.

Normier, R. 1977: Idg. Konsonantismus, germ. »Lautverschiebung« undVernersches Gesetz, KZ 91, p. 171-218.

Pokorny, J. 1948-1959: Indogermanisches etymologisches Wörterbuch I,Bern.

Prokosch, E. 1939: A comparative Germanic grammar, Philadelphia,Reichert, H. 1993: GUTANI ? WI HAILAG, Sprache 35, 1991-1993, p. 235-

247.ROSS, A. S. C., and J. Berns 1992: Germanic, in: Indo-European numerals,

ed. by J. Gvozdanovic, Berün, New York (Trends in Linguistics Studiesand Monographs 57), p. 555-715.

Schirlitz, S. Ch. 1908: Griechisch-deutsches Wörterbuch zum Neuen Testa-ment. 6th ed., Giessen.

Schwyzer, E. 1939: Griechische Grammatik I, München.Speirs, A. G. E. 1984: Proto-Indo-European laryngeals and ablaut, Amster-

dam.Stang, C. S. 1966: Vergleichende Grammatik der baltischen Sprachen, Oslo,

Bergen, Troms0.Steller, W. 1928: Abriss der altfriesischen Grammatik, Halle (Saale).Streitberg, W. 1920: Gotisches Elementarbuch. 5th/6th ed., Heidelberg.- 1928: Die gotische Bibel II: Gotisch-griechisch-deutsches Wörterbuch.

2nd ed., Heidelberg.1950: Die gotische Bibel I: Der gotische Text und seine griechischeVorlage. 3rd ed., Heidelberg.1963: Urgermanische Grammatik. 3rd ed., Heidelberg.

Suzuki, S. 1992: Toward an explanatory account of Thurneysen's law inGothic, PBB 114, p. 28-46.1994: Final devoicing and elimination of the effects of Verner's law inGothic, IF 99, p. 217-251.

Thurneysen, R. 1898: Spirantenwechsel im Gotischen, IF 8, p. 208-214.1946: A grammar of Old Irish, rev. and enl. ed., tr. by D. A. Bincheyand O. Bergin, Dublin.

Vasmer = Fasmer, M. 1986-1987: Etimologiöesky slovar' russkogo jazyka,tr., äug. and rev. by O. N. TYuba£ev. 2nd ed., 4 vols., Moscow.

Wackernagel, J., and A. Debrunner 1954: Altindische Grammatik II.2: DieNominalsuffixe, Göttingen.

Walde, A., and J.B. Hofmann 1838-1854: Lateinisches etymologischesWörterbuch. 3rd ed., 2 vols., Heidelberg.

Whitelock, D. 1974: Sweet's Anglo-Saxon reader. 15th ed., 2nd rev. imp.,Oxford.

Woodhouse, R. 1995: Proto Indo-European injective asperes [sie], IF 100,p. 92-100.1997: Merlingen's implosives, Bartholomae's law, Germanic aspiratesand consonant shifts, PBB 119, p. 374-398.1998: Verner's and Thurneysen's laws in Gothic äs evidence for ob-struent development in early Germanic, PBB 120, p. 194-222.

Wright, J. 1907: Historical German grammar I: phonology, word-formationand accidence, Oxford.




1954: Grammar of the Gothic language. 2nd ed. with a Supplement tothe grammar by O. L. Sayce, Oxford.

Wright, J., and E. M. Wright 1925: Old English grammar. 3rd ed., Oxford.BRISBANE/QUEENSLAND ROBERT WOODHOUSE



the origin of thurneysen's law

Documents